Best MCP Servers (2026)

Amazon Bedrock AgentCore is AWS's dedicated enterprise runtime for securely deploying and operating AI agents at any scale without infrastructure management. Provides complete session isolation, Amazon VPC connectivity, AWS PrivateLink support, managed memory, code execution, and observability. Complements Bedrock Agents with production-grade security controls. Usage-based pricing.

Amazon Bedrock Agents is AWS's managed service for building and deploying enterprise AI agents on top of foundation models. Agents orchestrate multi-step tasks using APIs, enterprise data sources, and action groups with no infrastructure management. AgentCore adds enterprise-grade isolation, VPC connectivity, PrivateLink, and session management. Used by 100,000+ organizations. Usage-based pricing.

Manage Docker containers and images via MCP. Start, stop, inspect, and build containers from agent workflows.

LiveKit is a real-time audio and video infrastructure platform and the foundation on which LiveKit Agents is built. It provides a global WebRTC media server network, server-side SDKs in Python, Go, Node.js, and more, plus browser and mobile client SDKs. For voice agent use cases, LiveKit handles multi-party audio rooms, direct phone integration via SIP/PSTN, and real-time media transport with <100ms latency. LiveKit Cloud offers managed hosting with usage-based pricing; the entire server stack is open source and self-hostable for teams that want full control over their media infrastructure.

Developer-first identity platform providing authentication, authorization (RBAC, SCIM), MFA, SSO, and fraud detection for applications that need to authenticate both human users and machine/agent identities. Backend and client-side SDKs support fine-grained API access control. Free for the first 10,000 MAUs; $0.10/user/month after that.

Connects AI agents to Datadog for monitoring metrics, logs, and alerts through MCP.

Monday.com work OS via MCP. Boards, items, columns, automations.

Official Stripe MCP server. Query customers, charges, subscriptions, and products. Full read/write access with careful permission scoping.

Customer messaging platform via MCP. Conversations, contacts, articles, events.

Cloudflare's managed MCP servers offer robust protocol support with strong enterprise backing, excellent sandboxing, and official docs, tempered by limited performance data and a vendor-related incident.

Search API built for AI agents. Returns structured, citation-ready results. Faster and more agent-friendly than general search APIs.

Official Linear MCP integration. Create issues, manage projects, query team workflows. Clean API surface with good error handling.

Official Supabase MCP server excels in protocol support and security scoping but lacks performance data and full stability due to pre-1.0 versioning.

Durable execution for agents. Start workflows, signal running executions, query state. Best-in-class reliability for long-running agent tasks.

Official GitHub MCP server excels in MCP interop and trust but carries prompt injection risks requiring careful token scoping.

ThoughtSpot's Spotter is an enterprise agentic analytics platform with specialized AI agents: SpotterModel for governed semantic modeling, SpotterViz for automated dashboard generation, and SpotterCode for AI-assisted embed coding. Integrates with Snowflake, BigQuery, Redshift, and other warehouses. SpotCache enables unlimited AI analytics at fixed cloud cost. API-first with embed SDK.

Miro whiteboard via MCP. Create sticky notes, shapes, connectors, and read board content.

PayPal payment operations via MCP. Create orders, capture payments, manage subscriptions.

Vertex AI Agent Builder is Google Cloud's full-stack enterprise platform for building, deploying, and governing AI agents in production. Supports open frameworks like ADK and LangChain, with managed Agent Engine runtime, Sessions, Memory Bank, and Code Execution. Provides audit trails, granular IAM permissions, and secure-by-design compliance. Usage-based pricing per API call and compute.

Serverless MySQL via MCP. Branch databases, deploy schema changes, query with connection pooling.

Background jobs and scheduled tasks via MCP. Reliable job queuing for agent-triggered async work.

Webflow CMS and site management via MCP. Query collections, create items, manage pages.

Mature hosted MCP server with strong official docs, token-based security, and enterprise backing, ideal for real-time data agents but lacks detailed failure recovery and load testing evidence.

Prisma ORM operations via MCP. Introspect schemas, run queries, manage migrations from agent workflows.

Stripe's official MCP-oriented toolkit for agent-triggered payments and billing operations with approval controls.

Enables AI agents to query dashboards, metrics, and visualizations from Grafana via MCP.

Financial data access via MCP. Account balances, transactions, identity verification. High security bar.

Azure AI Foundry Agent Service is a managed, pro-code enterprise runtime for deploying, orchestrating, and governing AI agents at scale. Connects models, tools, and frameworks into a single secure runtime with enterprise identity, VPC networking, content safety, and audit observability. Used by 80,000+ enterprises and 80% of Fortune 500 companies. Billed on Azure consumption.

Product analytics via MCP. Query events, feature flags, session recordings, experiments.

Trello board management via MCP. Cards, lists, boards, attachments.

Official MongoDB Atlas MCP server. CRUD operations, aggregation pipelines, full-text search from agent workflows.

LiveKit Agents is an open-source framework and managed cloud platform for building real-time voice, video, and multimodal AI agents. The SDK handles the hardest parts of voice AI—streaming STT-LLM-TTS pipelines, turn detection, interruption handling, and telephony integration—with plug-in support for all major AI providers. Developers can self-host the entire stack on LiveKit Server (the most-used open-source WebRTC server) or deploy to LiveKit Cloud. The framework is free and open source; LiveKit Cloud charges usage-based rates for hosted inference and media routing with a free tier for development.

Serverless Postgres via MCP. Branch databases, query with natural language, manage schemas. Good fit for agent-driven development workflows.

Official Shopify MCP server. Products, orders, customers, inventory management from agent workflows.

Atlassian's official Remote MCP Server delivers robust MCP protocol support with enterprise-grade security and reliability, ideal for agentic AI integrations with Jira/Confluence, though beta status limits some maturity scores.

AWS NLP service with dedicated PII detection APIs (Detect PII, Contains PII) for identifying and redacting sensitive entities in text at scale. Supports real-time and batch processing. Used to prevent agent pipelines from logging or transmitting sensitive user data. Free tier: 50,000 units/month per API for 12 months. Then usage-based per 100 characters.

Linear's official MCP server delivers solid MCP protocol support with strong trust signals from a mature company, though lacks performance data and full security scoring due to profile.

Secure cloud sandbox infrastructure for running AI-generated code in isolated microVM environments. Agents execute untrusted code without risk to host systems. Supports Python, JavaScript, and other runtimes. SDK integrates with any LLM. Usage-based billing per second of sandbox runtime (~$0.05/hr for 1 vCPU). Free Hobby plan includes $100 credit; Pro at $150/month.

Google Gemini Enterprise is an enterprise intranet search, AI assistant, and agentic workflow platform (formerly Google Agentspace). Enables knowledge workers to discover, create, share, and run AI agents grounded in organizational data. Integrates with Google Workspace, third-party SaaS, and internal data sources via APIs. Available through Google Cloud with custom enterprise pricing.

Atlassian Confluence knowledge base via MCP. Search pages, create content, manage spaces.

ClickUp task management via MCP. Tasks, spaces, lists, goals, and time tracking.

Identity-based secrets management platform for securely storing, distributing, and governing access to API keys, tokens, and credentials used by AI agents. Fully API-driven with dynamic secrets, automatic rotation, and fine-grained access policies. Open-source self-hosted version is free; Vault Secrets SaaS free up to 25 secrets; Enterprise pricing custom.

Firecrawl is a web data API that turns any website into clean, LLM-ready Markdown or structured JSON, powering AI agents and RAG pipelines with real-time web content. It handles JavaScript rendering, CAPTCHA solving, proxy rotation, and concurrent crawling behind a simple API. An AI Extract feature pulls structured data from pages using a plain-English prompt. A free tier offers 500 one-time credits; paid plans start at $16/month for 3,000 credits, scaling to $333/month for 500,000 credits.

AWS serverless compute platform for running agent logic, inference tasks, and event-driven AI workflows without provisioning servers. Supports Python, Node.js, and container images for packaging LLM inference with ONNX or lightweight models. Billed per request and per GB-second of execution. Free tier: 1M requests and 400K GB-seconds monthly. Popular for cost-efficient AI inference when GPU is not required.

Anchor is a cloud-hosted browser platform that lets AI agents interact with the web like a human, running up to 50,000 concurrent sessions in the cloud or on-premises. It is designed to be deterministic where possible — reverting to AI reasoning only when required at runtime — making it more reliable than fully LLM-driven approaches. Anchor has partnered with Cloudflare for verified browser agent support and is trusted by Groq, Unify, and Browser-use. Pricing is $0.01 per step with $5 in free monthly credits.

MCP server connector for Perplexity API enabling web search. Allows agents to perform research without leaving MCP ecosystem.

Cohere North is a privacy-first enterprise AI agent platform deployable in private VPCs, on-premises, or via Cohere's secure Model Vault. Enables design and deployment of agent-powered workflows connected to enterprise tools via flexible APIs and built-in connectors. Features granular access control, continuous red-teaming, and compliance with GDPR, SOC-2, and ISO 27001. Subscription-based.

Playwright is Microsoft's open-source browser automation framework supporting Chromium, Firefox, and WebKit with a single API across JavaScript/TypeScript, Python, Java, and .NET. It is the most widely adopted foundation for building browser-based AI agents, used directly by Stagehand, AgentQL, and dozens of agent frameworks. Playwright now includes native AI agent support via the MCP server and integrates with GitHub Copilot's Coding Agent. It is entirely free and open-source under the Apache 2.0 license.

Real-time web search with citations via MCP. Research-grade search with source attribution.

Kubernetes cluster management via MCP. Deployments, pods, services, config maps. High capability, high risk — approval mode required.

Allows AI agents to access Figma files, designs, and components through MCP protocol.

Popular open-source MCP memory server with strong protocol support and performance but limited formal docs and enterprise trust signals.

Cloud browser automation for agents. Managed Chrome instances with session replay, anti-detection, and proxy support.

Enables AI agents to access Cerebras code generation and inference capabilities via MCP.

AssemblyAI provides production-ready speech-to-text and audio intelligence APIs used widely as the STT layer in voice agent stacks. Its Universal model supports both pre-recorded and real-time streaming transcription with speaker diarization, sentiment analysis, entity detection, and topic classification available as add-ons. The streaming STT API is purpose-built for low-latency agent pipelines with sub-500ms transcript delivery. Pricing starts at $0.15/hr (Universal) with $50 in free credits; streaming audio billed identically to batch, with audio intelligence features priced separately per hour of audio processed.

Official Pinecone vector database MCP. Upsert and query embeddings directly from agent pipelines. Clean schema, reliable uptime.

Connects AI agents to Jira, Confluence, and Bitbucket for project management via MCP.

MCP server for querying and analyzing MongoDB collections. Provides agent builders with database access tools for NoSQL data operations.

Official Figma MCP server that feeds design tokens, component variants, text layers, and component properties directly to AI coding assistants including GitHub Copilot, Cursor, and Claude Code. Transforms Figma design files into machine-readable infrastructure. Supports two-way sync with Claude Code — publish designs from Claude to Figma AND pull designs into Claude. Part of Figma's broader Code Connect ecosystem bridging visual and implementation layers.

Provides AI agents access to Firebase services including Firestore, Auth, and Functions via MCP.

Jina Reader is a free API that converts any URL to clean, LLM-friendly Markdown with a simple prefix: r.jina.ai/[URL]. It natively handles PDFs, images (via automatic captioning), and web pages across 29 languages, powered by the ReaderLM-v2 1.5B parameter model that supports documents up to 512K tokens. It is the fastest way to give an AI agent web reading capability without building scraping infrastructure. The Reader API is free with flexible rate limits; higher-volume usage is available via Jina's token-based billing.

Behavioral analytics via MCP. Query user journeys, funnels, retention cohorts.

Airtable bases via MCP. Read records, create rows, update fields, manage views. Good for low-code agent workflows.

Make.com automation scenarios via MCP. Trigger scenarios, pass data, monitor runs.

Zapier MCP Server delivers robust MCP protocol support with massive app ecosystem and strong security, tempered by limited technical docs and recent minor incident.

Twilio Programmable Voice is the industry-standard telephony API for building voice AI agents that make and receive phone calls over PSTN and VoIP. It supports TwiML (XML-based call control), WebSockets for streaming audio to AI models in real time, IVR flows, conference calling, call recording, and ASR in 120 languages. Twilio is the underlying telephony layer used by many voice AI platforms including Retell AI and Vapi. PAYG pricing is $0.013/min outbound and $0.0085/min inbound (US); phone numbers start at $1.15/mo. Volume discounts and committed-use pricing are available for enterprise.

Incident management via MCP. Trigger alerts, acknowledge incidents, manage escalations.

Pipedream MCP excels in MCP protocol support and massive integrations with strong security/compliance, though lacks performance data and detailed error docs.

Unified inference platform for packaging, serving, and scaling AI models and multi-model pipelines in Python. Supports any model format and runtime, with built-in task queues, dynamic batching, multi-GPU orchestration, and distributed serving. BentoCloud provides managed compute for rapid production deployment. Used by agent builders to compose and serve LLMs, embeddings, and custom models as microservices.

Trigger n8n workflows via MCP. Execute webhooks, pass data to automation pipelines, retrieve results.

Official Vercel MCP server. Deploy projects, manage environment variables, inspect deployments and build logs.

Official Pinecone MCP server excels in protocol support and docs for dev tools but limited by early access status and lack of production readiness evidence.

Trigger and monitor GitHub Actions workflows via MCP. List runs, check status, download artifacts.

Open-source LLM proxy and SDK that routes agent calls to 100+ LLM providers through a single OpenAI-compatible API. Handles cost tracking, load balancing, rate limiting, guardrails, and spend controls across Anthropic, OpenAI, Bedrock, Azure, and more. Self-hostable for free; enterprise tier adds SSO, audit logs, and Prometheus metrics. Critical infrastructure for multi-provider agent cost management.

Serverless cloud platform that runs Python functions in isolated containers using gVisor sandboxing for secure execution of untrusted code. Agent builders use Modal to run AI-generated code without exposing host systems. Sub-second cold starts with per-second billing. Free tier includes $30/month compute credits; GPU compute from $0.59/hr for T4.

Vapi is a developer-first voice AI infrastructure platform for building, testing, and deploying voice agents at scale. It provides an abstraction layer over STT, LLM, and TTS providers, letting developers swap in any combination of services via a single API. The platform handles real-time call orchestration, WebSocket streaming, tool calling, and structured output. Pricing is $0.05/min for Vapi's platform layer, with LLM, voice, and telephony billed separately from third-party providers; total production costs typically run $0.25–$0.33/min.

Bright Data is an enterprise-grade web data infrastructure platform providing proxies, web scraping APIs, and datasets for AI development. It offers 150M+ real-user IPs across 195 countries, pre-built scrapers for 120+ domains, a Scraping Browser for complex JavaScript-heavy pages, and a Web Unlocker for bypassing anti-bot systems. The platform powers AI model training, RAG pipelines, and agentic web access for 20,000+ companies. Pricing is usage-based starting at $0.001/record; datacenter proxies from ~$10.50/GB.

Unified API gateway providing access to 300+ AI models from 60+ providers through a single OpenAI-compatible endpoint. Routes agent model calls with automatic provider fallback, latency-optimized routing, and cost comparison. Charges at provider rates with no markup and no subscription required — buy credits and pay as you go. Critical for agent systems requiring model diversity, redundancy, and cost optimization across providers.

Strong MCP server leveraging Neon's stable infrastructure with excellent protocol support but limited performance and audit details.

Connects AI agents to Supabase for database queries, authentication, and edge functions through MCP.

Reliable hosted MCP protocol server from established Notion with solid interop and docs, lacking detailed error/performance evidence.

Deepgram provides a high-accuracy, low-latency speech-to-text API built for production voice AI applications. Its Nova-3 model delivers real-time streaming transcription at $0.0077/min and batch transcription at $0.0043/min, with $150 in free credits to start. Beyond transcription, Deepgram offers text-to-speech, speaker diarization, sentiment analysis, and a Voice Agent API that bundles STT, LLM routing, and TTS into a single WebSocket session. The platform is widely used as the STT backbone inside Retell AI, Vapi, and Pipecat pipelines.

Asana task management via MCP. Create tasks, manage projects, update assignees.

Sora is OpenAI's video generation model that produces cinematic-quality videos up to 25 seconds from text prompts. The API (Sora 2) is available through OpenAI's developer platform with usage-based pricing: $0.10/second for 720p standard and up to $0.50/second for 1024p Pro quality. A minimum $50 credit purchase is required for API access. ChatGPT Plus ($20/mo) and Pro ($200/mo) subscribers also get direct web access. Essential for AI builders who need OpenAI-ecosystem video generation with tight integration into existing GPT-based workflows.

ElevenLabs is a leading voice AI platform offering ultra-realistic text-to-speech, voice cloning, and a fully managed Conversational AI API. Developers can clone any voice from a short sample, stream audio with sub-200ms latency, and build end-to-end voice agents using the EVI (Empathic Voice Interface) SDK. The Conversational AI product handles STT, LLM turn management, and TTS in one hosted pipeline with support for tool calling and interruptions. Pricing ranges from a free tier (10,000 chars/mo) up to a $1,320/mo Business plan; API and enterprise pricing is usage-based by character.

Centralized cloud secrets manager that syncs environment variables and API keys to agent infrastructure across environments. Provides a REST API for programmatic secret access, supports service accounts for machine identities, and integrates with CI/CD pipelines. Free plan for up to 3 users; Team plan at $21/user/month with no extra cost for service accounts.

Open-source AI-native vector database supporting hybrid dense and sparse search for agent memory, RAG, and knowledge retrieval. Deployable as self-hosted Docker/Kubernetes or via Weaviate Cloud (shared/dedicated). Integrates natively with LangChain, LlamaIndex, OpenAI, and Cohere. Serverless Cloud starts at $25 per 1M vector dimensions/month. Supports multi-tenancy for serving many agent instances from one cluster.

Managed vector database purpose-built for production AI and agent applications. Stores and retrieves high-dimensional embeddings for RAG, semantic search, and agent long-term memory with sub-millisecond query latency at billion-vector scale. Includes Pinecone Assistant for agent-based chat, and Pinecone Inference for managed embedding models. Free Starter tier; Standard plan from $50/month minimum.

Durable workflow triggers via MCP. Send events, cancel runs, query function history. Excellent for agent orchestration.

Cloud API platform for running and deploying AI models without managing infrastructure. Hosts 100+ official models — LLMs, image generation, audio — always on with stable APIs, plus custom model deployment via Cog. Scales automatically from zero to high traffic. Agents call models via simple REST API. Billed per-second of compute only when models are running; no idle charges.

Official Cloudflare MCP server. Manage Workers, KV, R2, D1, and Durable Objects from agent workflows.

Official MCP server for Brave Search API. Web and local search with privacy-first results. Reliable, fast, no rate limit surprises on paid plans.

Composio delivers robust agent tool integrations with strong interop and security but lacks latency data and explicit data portability.

Anthropic's Computer Use is a capability that lets Claude control a desktop environment — moving cursors, clicking buttons, typing text, and taking screenshots — enabling fully autonomous computer-use agents. It is accessible via the Claude API as a tool alongside standard message APIs, with no additional setup beyond standard tool-use billing. Developers can build agents that automate any GUI task across any application or website. Pricing follows standard Claude API token rates (e.g., Sonnet 4.6 at $3/M input, $15/M output), with a small ~480-token system prompt overhead per call.

Self-hosted OpenAI-compatible proxy server from LiteLLM that centralizes LLM access for all agents in a deployment. Manages virtual API keys per agent or team, enforces per-key spend budgets and rate limits, load-balances across multiple providers, and logs all requests for cost attribution. Prevents runaway agent costs and simplifies multi-provider LLM infrastructure. Free to self-host; enterprise from $250/month.

ZenRows is a web scraping toolkit that handles rotating proxies, anti-bot bypassing, fingerprinting, CAPTCHA solving, and JavaScript rendering behind a single API. It offers a Universal Scraper API, a Scraping Browser for Puppeteer/Playwright workflows with built-in anti-detection, and 55M+ residential proxies with geotargeting. A free trial covers 1,000 URLs with no credit card required. Paid plans start at $69/month for 250,000 credits; JavaScript rendering with premium proxies uses 25 credits per request.

Real-time API security layer for AI agents that detects prompt injections, jailbreaks, indirect prompt injection, PII leakage, toxic content, and malicious links. Integrates in under 5 minutes via a single API call wrapping any LLM call. Available as SaaS or self-hosted. Free tier available; enterprise and custom plans via demo request.

Qdrant vector database MCP server. High-performance similarity search for agent memory and RAG systems.

Open-source LLM engineering platform providing traces, evaluations, prompt management, and metrics. Supports LLM-as-judge scoring, human feedback collection, manual labeling, and custom evaluation pipelines via API. Integrates with OpenAI, LangChain, LlamaIndex, and LiteLLM via OpenTelemetry. Cloud-hosted freemium with paid tiers from $29/month; self-hostable under MIT license.

High-performance open-source vector database and similarity search engine written in Rust, designed for billion-scale agent memory and semantic retrieval. Deployable via Docker, Kubernetes Helm chart, or Qdrant Cloud (managed). Supports ANN indexes, filtering, multi-vector payloads, and distributed horizontal scaling. Official client SDKs for Python, TypeScript, Go, Java, and Rust. Free Cloud tier available.

Cloud-native open-source vector database built for massive-scale similarity search, maintained under LF AI & Data Foundation with Apache 2.0 license. Supports Lite (Python in-process), Standalone (single Docker), and Distributed (Kubernetes for billion-scale) deployment modes. Integrates with LangChain, LlamaIndex, OpenAI, and HuggingFace for RAG and agent memory pipelines. Managed via Zilliz Cloud.

Apify is a full-stack web scraping and automation platform with a marketplace of 10,000+ ready-made automation tools called Actors — serverless cloud programs that scrape data, automate web tasks, and run AI agents. It provides an MCP server so AI models like Claude can dynamically discover and use Actors as tools. Apify is trusted by Intercom, the European Commission, and thousands of developers, with SOC 2 Type II compliance. Free tier includes $5/month in credits; paid plans start at $29/month.

IBM watsonx Orchestrate is an enterprise multi-agent orchestration platform offering 100+ domain-specific AI agents and 400+ prebuilt tools. Supports both no-code drag-and-drop and pro-code agent construction with IBM Granite reasoning models. Integrates with enterprise HR, finance, procurement, and sales apps. Governed with centralized oversight, built-in guardrails, and automated policy enforcement.

Hyperbrowser is cloud browser infrastructure for AI agents offering sub-second session launch times and support for 1,000+ concurrent sessions without latency degradation. It integrates with OpenAI's CUA and Claude Computer Use, includes built-in stealth features and automatic CAPTCHA solving, and provides Python and Node.js SDKs. Hyperbrowser is well-suited for large-scale, high-concurrency agentic workloads. Pricing is credit-based: one browser hour costs ~$0.10 and one scraped page costs ~$0.001.

Official browser automation MCP using Puppeteer. Navigate pages, fill forms, take screenshots, extract content.

AI observability and evaluation platform for production LLM systems. Runs experiments against datasets with automated scorers (LLM-as-judge, factuality, code-based), tracks regressions in CI/CD via GitHub Actions, and provides real-time production tracing. Multi-language SDKs (Python, TypeScript, Go, Ruby, C#). SOC 2 Type II, HIPAA, GDPR compliant. Free tier with paid plans from $249/month.

AI-native data loss prevention platform with a developer API for detecting PII, PHI, secrets, and credentials in text, files, and cloud data sources. Agents integrate via REST API to scan outputs before storage or transmission. Free plan covers 3 GB/month scanned; Enterprise plans offer custom volume limits and SLA support.

Provides AI agents with financial market data, stock quotes, and technical indicators via MCP.

HeyGen is a leading AI avatar video generation platform enabling developers and marketers to create talking-head videos from text scripts using customizable AI avatars. Its developer API is billed by the minute (1 credit = 1 minute of avatar video), with plans starting from $5 for API credits. The Scale plan drops unit cost to $0.50/minute and unlocks translation and proofreading APIs. HeyGen supports video translation into 40+ languages and provides a full REST API for programmatic video generation at scale, making it the go-to choice for AI-driven video localization and personalization products.

Dream Machine by Luma AI is a text-to-video and image-to-video generation platform known for its cinematic motion quality and photorealistic output. It supports API access through a separate credit system from web subscriptions, with web plans from $9.99/month (Lite) to $94.99/month (Unlimited). A 5-second video costs 170 credits and a 10-second video costs 340 credits on web. Developers building AI creative tools can access Dream Machine programmatically; the Plus plan ($29.99/mo) is the minimum for commercial use.

Allows AI agents to crawl websites and extract structured data via MCP server.

MCP server for natural language task management in Todoist. Enables agents to create, update, and query tasks programmatically.

CNCF-graduated open-source general-purpose policy engine for enforcing authorization rules across microservices, APIs, Kubernetes, and agent systems. Developers write policies in the Rego language and query OPA via REST API or Go library to make access control decisions for agent tool calls and resource access. Free and open source.

Cognigy.AI is an enterprise-grade conversational AI platform for building voice and chat agents deployed in contact center and customer service environments. It provides a low-code flow builder, a generative AI copilot layer, NLU, live agent handoff, omnichannel routing (voice, chat, messaging), and extensive CRM and CCaaS integrations. The platform runs on-prem or in the cloud and supports compliance requirements for regulated industries. Pricing starts at approximately $2,500/mo for enterprise subscriptions, with average annual contracts around $115,000; no self-serve or free tier is available.

Speechmatics is a multilingual speech recognition API supporting 55+ languages and 69 translation pairs, designed for enterprise voice AI workloads requiring high accuracy across diverse accents and dialects. The API offers real-time streaming and batch transcription, speaker diarization, punctuation, and an enterprise real-time STT model with sub-second latency. It targets applications in contact centers, media, and voice agent post-call analytics. Pricing includes a free tier (8 hrs/mo, monthly reset), PAYG at approximately $0.0117/min, with automatic volume discounts above 500 hours; enterprise customers receive custom negotiated rates.

Offers AI agents access to Jina AI's search, embedding, and multimodal tools via MCP.

Enables AI agents to perform file system operations like reading, writing, and listing files through a secure MCP server.

Durable execution platform for building reliable, fault-tolerant agent workflows and long-running AI pipelines. Automatically captures workflow state at every step — agents survive failures, retries, and infrastructure restarts without losing progress. MIT-licensed open source for self-hosting; Temporal Cloud managed service with $1,000 free credits on signup. Widely used for multi-step agent orchestration that must be resilient to chaos.

MCP server providing access to Redis key-value stores through standardized tools. Enables fast caching and session storage for agents.

Distributed SQLite via MCP. Create databases, run queries, manage schemas at the edge. Good for multi-tenant agent workflows.

Developer-friendly email via MCP. Clean API, React email templates, excellent deliverability. Better DX than legacy email APIs.

Agentforce is Salesforce's enterprise agentic AI platform that orchestrates autonomous agents across sales, service, and employee workflows 24/7. Developers build and deploy agents using Agentforce Builder with MuleSoft connectors, REST APIs, and the Atlas reasoning engine. Supports the full agent lifecycle—build, test, deploy, monitor. Pricing via Flex Credits, conversation-based billing, or per-user licensing.

Browserbase is a cloud browser infrastructure platform purpose-built for AI agents, providing managed headless browsers that are fast, reliable, and scalable. It integrates natively with Playwright, Puppeteer, Selenium, and its own Stagehand framework via the Chrome DevTools Protocol. Features include isolated browser sessions, CAPTCHA solving, stealth proxies, and SOC-2 Type 1 and HIPAA compliance. Plans start with a Free tier (1 concurrent browser, 1 hr/mo), Developer at $20/mo, and Startup at $99/mo.

Fine-grained authorization-as-a-service platform with native support for AI agent identity and access control. Assigns machine identities to agents, enforces permissions per tool and resource, and integrates with LangChain, PydanticAI, and MCP. Supports RBAC, ABAC, and ReBAC policies. Free tier for small teams; Pro and Enterprise tiers based on monthly active identities.

The reference MCP implementation with excellent protocol support, dragged down by two 2025 CVEs and missing server-level changelog.

Serverless GPU infrastructure platform for AI inference and training with sub-10-second cold starts and instant autoscaling. Python-native API to define and deploy agent tasks, background jobs, and LLM inference pipelines across multi-cloud GPU fleets including H100s and 4090s. Open-source engine (Beta9) supports bring-your-own cloud. Scale-to-zero billing eliminates idle GPU costs for infrequent agent workloads.

Notte is a browser infrastructure platform for AI agents that provides remote browser sessions, serverless automation functions, authentication primitives, and a visual studio. It is model-agnostic, supporting OpenAI, Anthropic, Gemini, and others, and enables agents to navigate, extract data, and complete tasks at scale. Notte raised $2.5M pre-seed and is designed for production-grade agentic automation. A free tier includes 100 browser hours; paid plans start at $20/mo plus usage.

Open-source microVM monitor from AWS that provides KVM-based hardware-level isolation for untrusted workloads. Boots microVMs in 125ms with under 5MB memory overhead; supports 150 VMs/sec/host. Powers AWS Lambda and AWS Fargate. Developers use it to build agent sandboxing infrastructure where AI-generated code runs in fully isolated environments. Free, Apache 2.0 licensed.

Gladia is an audio transcription and intelligence API built for real-time and async speech processing in agent pipelines. It supports multilingual transcription, speaker diarization, live streaming, and audio intelligence features like named entity recognition and summarization—all bundled into a single per-hour rate without add-on fees. The API handles pre-recorded and live audio with a unified interface, making it popular for meeting intelligence and voice agent post-call analytics. Pricing starts free (10 hrs/mo), with PAYG at $0.20/hr async and $0.25/hr real-time; enterprise plans include custom models and fine-tuning.

Open-source multi-type data labeling platform for building evaluation datasets and ground-truth sets for AI systems. Supports text, images, audio, video, and time-series annotation with ML-assisted labeling, configurable templates, and inter-annotator agreement scoring. Used by NVIDIA, Meta, and Cloudflare. Free open-source edition; enterprise plans via HumanSignal with cloud trial.

Query Sentry errors, performance data, and releases. Useful for agents involved in incident response.

Fully managed serverless container platform on Google Cloud that natively supports hosting AI agents built with ADK, LangGraph, Dify, and other frameworks. Auto-scales from zero to handle traffic spikes with per-100ms billing. GPU support for serverless ML inference billed per-second. Official Google documentation covers agent deployment patterns. Free tier includes 180K vCPU-seconds and 2M requests/month.

Knowledge graph-based persistent memory system via MCP. Allows agents to store, retrieve, and query long-term memory across sessions.

D-ID is an AI talking head video platform that animates still photos and generates realistic lip-synced videos using AI avatars driven by text or audio scripts. Its developer API offers per-second pricing at $0.40/second (standard) and $0.15/second (fast model), with a free 14-day trial including 3 minutes. The API Build plan starts at $18/month and deducts usage from the same balance as web studio usage. D-ID is widely used in conversational AI apps, AI tutoring platforms, and personalized marketing automation pipelines.

Scrapfly is a web scraping API that combines anti-bot bypass (Cloudflare, DataDome, PerimeterX, Akamai, and 20+ systems), CAPTCHA solving, and cloud browser rendering into a single endpoint. It offers 130M+ rotating proxies from 120+ countries, AI-powered data extraction with LLMs, and screenshot capture. Scrapfly is designed for developers building data pipelines and AI agents that need reliable, high-volume web access. Pricing is credit-based with prepaid monthly plans plus pay-as-you-go overflow; a free tier is available.

Zendesk's Resolution Platform provides enterprise-grade AI agents for customer and employee service, capable of resolving 80% of support issues autonomously via voice, chat, and email. Integrates with external knowledge sources (Confluence, SharePoint, Google Drive) via Knowledge Connectors. Exposes REST APIs for custom agent configuration and workflow integration. Used by ~20,000 enterprises.

Scheduling automation via MCP. Create booking links, check availability, manage event types.

Together AI is an AI infrastructure cloud that recently expanded into video generation, offering API access to models including MiniMax Hailuo, Google Veo 3.0, and OpenAI Sora 2 via OpenAI-compatible endpoints. Pricing uses per-model usage-based billing with transparent rates listed on the model pages. Together AI also hosts open-source models including Wan and Stable Diffusion variants. Existing Together AI customers can access video generation without changing their API integration pattern, making it a convenient choice for teams already using the platform for LLM or image generation workloads.

Browser Use is an open-source Python library that gives AI agents full control of a web browser, letting LLMs autonomously navigate, click, type, and extract data without pre-written scripts. It supports vision models (screenshot-based) and DOM extraction, and works with OpenAI, Anthropic, Google, and open-source models. The library has crossed 50,000 GitHub stars and is one of the fastest-growing AI open-source projects. The core library is free (MIT); Browser Use Cloud offers token-based pricing with $10 in free credits for new users.

AWS managed service for adding configurable content safety controls to AI agents, including content filters, denied topic classification, PII redaction, prompt attack detection, and hallucination checking. Works with any foundation model including non-AWS models. Blocks up to 88% of harmful content per AWS benchmarks. Priced at $0.15 per 1,000 text units for content filters.

Production AI gateway and observability platform that routes agent LLM calls across 1,600+ models with load balancing, fallbacks, retries, guardrails, and cost governance. Integrates natively with LangChain, LangGraph, CrewAI, and OpenAI Agents SDK so all model calls inherit routing and spend controls automatically. Logs and traces every request. Open-source gateway; managed cloud with usage-based pricing per recorded request.

Agora provides real-time voice, video, and AI engagement SDKs used by developers to embed low-latency communications into applications. Its Conversational AI Engine API enables developers to add AI voice agents to live voice channels, processing turn detection and AI responses in real time within Agora's global media network. The RTC SDK handles voice calling at $0.99 per 1,000 participant minutes; the Conversational AI Engine is priced from $0.0099/min of interaction. A free tier covers the first 10,000 minutes of video calling monthly, with additional free minutes for real-time speech-to-text during development.

Pega Infinity is an enterprise agentic transformation platform that orchestrates AI agents within deterministic workflow guardrails using Pega Agentic Process Fabric. Supports MCP and A2A protocols for third-party agent integration, with full audit trails via Agent Tracer. Agents built through the AI Agent Rule framework execute governed, predictable outcomes across regulated industries. Cloud-based subscription.

Runway is a professional AI video generation and editing platform offering Gen-3 Alpha and Gen-4 models for text-to-video, image-to-video, and video-to-video transformation. It exposes a developer API billed at $0.01 per credit, with Gen-3 Alpha costing 10 credits/second and the faster Turbo variant at 5 credits/second. Subscription plans start at $12/month with monthly credit allotments. Runway is widely used by AI product builders who need a stable, high-quality video generation API with predictable per-second billing.

Stagehand is an open-source AI browser automation SDK built by Browserbase that adds natural language control to Playwright-based browser automation. It combines the precision of code with the flexibility of LLM-driven instructions, making browser agents reliable and maintainable. Stagehand v3 is significantly faster on iframe and shadow-root interactions than previous versions. The SDK itself is free and open-source; running it at scale requires Browserbase cloud infrastructure, which starts at $20/month.

Vonage (now part of Ericsson) offers a Programmable Voice API for building AI voice agents with WebSocket streaming, ASR in 120 languages, TTS, IVR, call recording, and real-time third-party AI integration. Its AI Studio provides a drag-and-drop builder for deploying voice AI virtual agents using Vonage's proprietary NLU engine without deep coding. Vonage recently partnered with AWS to support conversational agents via Amazon Nova Sonic and introduced MCP integration for AI-meets-programmable-communications workflows. Pricing is per-second usage-based starting at $0.014/min for developer voice features; free credits available on signup.

The world's most-downloaded open-source AI platform (30M+ monthly downloads) with a built-in evaluation module for LLMs, agents, and RAG systems. Provides 50+ metrics and LLM judges, dataset versioning for test cases, automated regression detection, and production monitoring. Framework and cloud agnostic under Apache 2.0; integrates with Databricks, AWS, Azure, and GCP.

Oxylabs is an enterprise web data collection platform offering 177M+ IP proxies, a Web Scraper API with domain-specific endpoints, and AI-powered OxyCopilot that auto-generates scraping code from natural language prompts. It supports AI agent stacks with structured data extraction, JavaScript rendering, and geotargeted data collection. Oxylabs charges only for successful requests (2xx/4xx status codes), giving predictable billing. Web Scraper API plans start at $49/month; proxies are usage-based with volume discounts.

Allows AI agents to deploy and manage static sites on EdgeOne Pages via MCP protocol.

LLM and AI agent observability platform by LangChain with built-in evaluation capabilities. Runs offline evals against curated datasets to catch regressions and online LLM-as-judge evals against live production traffic. Framework-agnostic Python, TypeScript, Go, and Java SDKs; supports self-hosted, BYOC, and managed cloud deployments. Free tier available with paid plans scaling by trace volume.