Frameworks & SDKs tools for MCP

Open-source AI agent framework that turns LLMs into persistent, proactive personal AI agents with multi-channel messaging (Signal, Telegram, Discord, WhatsApp), cron scheduling, memory systems, MCP integration, skill plugins, sub-agent spawning, and browser automation

Web search, scrape and crawl API for AI agents. Turns any website, including JavaScript-heavy pages, into clean LLM-ready markdown or structured JSON, with browser actions (click, scroll, type), PDF and DOCX parsing, an MCP server and an autonomous agent endpoint. Open source (AGPL-3.0) and self-hostable

Open-source browser agent (Python and TypeScript) whose loop injects serialized DOM state each step, exposes browser actions as registered tools, and runs the same code against a local Chrome or a cloud browser. It has grown from a minimal reference into a full framework (~70k lines of Python across agent, DOM, tools, MCP, and sandbox modules) — study it for production DOM-to-context serialization

by Anthropic - Anthropic's official, curated directory of high-quality Claude Code plugins, installable from within Claude Code

Open-source personal agent harness: desktop app and Rust library with built-in memory, token compression, and instant tool search, efficient enough to run hundreds of agents on a $10 server.

Start building LLM-empowered multi-agent applications in an easier way. github

TypeScript-native agent framework (from the Gatsby team) with 22K+ stars and 300K+ weekly npm downloads. Connects to 40+ providers through one standard interface, with built-in workflows, RAG pipelines, and agent orchestration. The @mastra/deployer handles serverless deployment, and the eval system supports LLM-as-judge out of the box. The strongest alternative to Vercel AI SDK for teams that need

The fast, Pythonic way to build Model Context Protocol servers 🚀

June 2026 harness-first redesign built around the Capability primitive: a single composable unit bundling instructions, tools, lifecycle hooks, and model settings. The split between a small stable core and a fast-moving pydantic-ai-harness lets capabilities graduate as they prove essential, while deferred loading keeps unused tools out of the context window

Expose pipelines from RocketRide, a self-hosted, open-source AI pipeline tool with multi-LLM support, as tools

Self-healing browser harness that connects an LLM directly to your real Chrome via CDP. The critical design decision: the agent itself writes missing helpers and domain skills into the harness during execution, so the scaffold improves every run rather than requiring manual updates. At ~1k lines across four core files, it is the clearest published demonstration of an editable harness that learns f

Rust - Open-source Agent OS built in Rust. 137K LOC. 14 crates. 1,767+ tests. Zero clippy warnings

Agent framework and applications built upon Qwen>=3.0, featuring Function Calling, MCP, Code Interpreter, RAG, Chrome extension, etc

YC-backed production harness that compiles multi-agent objectives into deterministic execution DAGs with state persistence, crash recovery, cost enforcement, and human-in-the-loop oversight. The most complete open-source reference for running agent workloads in production rather than demos

An open source TypeScript Framework for building AI agents with built-in LLM observability. github

Praison AI is a low-code, centralized framework leveraging AutoGen and CrewAI to simplify creating and orchestrating multi-agent systems for LLM applications, emphasizing customization and ease of human-agent interaction github | demo | website

A Go implementation of the Model Context Protocol (MCP), enabling seamless integration between LLM applications and external data sources…

AWS's open-source, model-driven agent SDK that treats the loop, tool binding, and guardrails as first-class primitives. Supports Bedrock, Anthropic, OpenAI, Gemini, and Ollama with native MCP, built-in observability, and multi-agent patterns — the missing AWS open-source harness framework alongside Bedrock AgentCore

Enterprise-grade, commercial-friendly agentic workflow platform for building next-generation SuperAgents. github

GPT Computer Assistant is an unofficial app that brings ChatGPT functionality to Windows and Linux, allowing for screen reading, microphone use, system audio interaction, clipboard management, script execution, and more github

TypeScript-native multi-agent orchestration that automatically decomposes a natural-language goal into a task DAG, parallelizes independent nodes, and synthesizes results. With only three runtime dependencies, built-in MCP support, token budgets, retries, and context compaction, it is the lightest production-grade harness layer for Node.js backends

canonical spec, SDKs, issues, proposals, and reference materials

🚀 2026 最系统的 AI Agent 速成指南｜智能体实战教程 · 完整学习路径 + 实战项目 + 面试题库 · 对标大模型应用开发工程师岗位 · 覆盖LangChain / LangGraph / Coze / Dify / MCP / skills / LLM / RAG / 提示词 · 企业级部署与微调 · 从0到企业级落地 + 从学习到上线项目 + 面试准备一体化

Desktop app for evals, RAG and fine-tuning

AG2 (from AutoGen creators) is an open-source programming framework for building AI agents and facilitating cooperation among multiple agents to solve tasks

kagent is a kubernetes native framework for building AI agents

Define, Prompt and Test MCP enabled Agents and Workflows

Integration gateway that links AI agents to 600+ tools via unified MCP/OAuth interface with built-in scaling and monitoring

The open-source managed agent platform: run durable AI agents on any model and any machine, from sandboxes to your own laptop, with MCP, skills, subagents, and human approvals. A self-hostable alternative to Claude Managed Agents and OpenAI's Agents API

A high-level framework for building MCP servers in TypeScript

A lightweight utility designed to simplify the deployment and management of MCP servers, ensuring ease of use, consistency, and security through containerization

Open-source production-ready RAG platform with built-in agentic reasoning, hybrid search, and multimodal support

Go framework for building production agent systems with graph workflows, tools, memory, A2A, AG-UI, MCP, evaluation, and observability

Local payment layer for handling x402 payment challenges with wallet-authorized stablecoin signing

Stop wrestling with countless APIs and complex integrations. AgentDock provides the open-source foundation to build, manage, and deploy production-ready AI agents and workflows, frictionlessly. github

Local-first, model-agnostic Agent Operation Environment (AOE) for building, borrowing, and orchestrating specialist agents and teams across supported LLM hosts, with owner-scoped packages and host-enforced permissions

Source-available, self-hosted AI workflow automation platform with a visual canvas for agent, RAG, and tool-using workflows. Includes MCP support, evals, traces, and cost tracking

Lightweight Python agent framework with memory, MCP/SSE integration, reusable Skills, Tree-of-Thought planning, OpenAI-compatible streaming, and LightSwarm multi-agent collaboration

Terminal coding agent and Python agent framework built on Pydantic AI, with planning, sub-agents, MCP and optional Docker-sandboxed execution; works with any model, including local ones

IBM Research's generalist agent harness built on LangGraph; executes tasks over APIs (OpenAPI specs, MCP servers, LangChain tools), the web (Playwright), or both, with policy-based guardrails and human-in-the-loop approval (Intent Guard, Playbook, Tool Approval), multi-agent supervision including remote A2A agents, and agent skills. Python, Apache-2.0

Bridge between Ollama and MCP servers, enabling local LLMs to use Model Context Protocol tools

One messaging identity for an AI agent across Slack, Discord, Telegram, Instagram, email, and X — a single on_message handler with threading, webhook verification, and platform quirks handled. Python + TypeScript SDK

Fast and elegant Typescript framework for building MCP servers

Autonomous agent framework that runs unattended on GitHub Actions, triggered by cron schedules or repository events. Dispatches skills to six coding-agent harnesses behind one contract (Claude Code, Grok, Codex, Pi, Vibe, Kimi), defines agent behavior as Markdown skills, persists memory in the git repository, evaluates its own run output to revise underperforming skills, and ships an MCP server ex

Local-first database and workspace for AI agents that keeps structured data, durable knowledge, documents, files, skills, and apps in one searchable place. Agents connect through MCP, OpenAPI, CLI, or coding-agent skills, and material writes can remain reviewable ChangeRequests before becoming canonical

Local-first managed agent runtime with persistent sessions, sandboxed execution, MCP tools, audit, replay, and a human console

Local-first, offline encrypted vector memory for AI agents over MCP or CLI, with AEAD-encrypted-at-rest records and embeddings, RAM-resident exact vector search, per-record crypto-shred deletion, and a hash-chained audit log. Apache-2.0, Python

Coordination protocol above Git: agents declare intent, semantic scope, and operation before writing code, deterministic rules raise findings when plans collide, and acceptance is gated on a verification run it executes itself. One Rust binary over local SQLite, with MCP for Claude Code, Codex, and Cursor.

Routes one coding thread across harnesses with quota-aware rotation between subscription profiles, Best-of-N runs, and cross-family review.

[Now with MCP Support] AI For Devs - Build, Share & Run agentic workflows. Just Docker. Just Markdown. BYO LLM

Separate signer that allows, denies or holds each agent tool call for a person before it runs and records every call in an Ed25519-signed, hash-chained log that verifies offline; Python and TypeScript SDKs, framework integrations, coding-agent hooks and an MCP gateway

A complete solution for agent memory, from extraction and vector search to automated optimization, with a REST API, MCP, CLI, and insights dashboard out-of-the-box

Shared context, memory, and task coordination as a single Go binary over local SQLite with hybrid keyword and semantic search.

MCP gateway that scopes agent tool access by identity, team, and role, with security policy enforced on each action and an audit trail. Useful as a reference for putting authorization at the tool boundary instead of relying on model prompts to restrict access

Open-source context and control plane for AI agents. Connect company tools once and share permission-aware context across ChatGPT, Claude, Codex, OpenClaw and custom agents, with MCP, scoped identities, approvals and audit

Universal CLI for AI agents — exposes the agent-runnable web, desktop, Electron, and bridge-CLI surface as deterministic commands. Declarative YAML adapters with structured error envelopes (adapter_path, step, suggestion) let agents edit failing adapters and retry. Optional MCP server (stdio + Streamable HTTP). Live catalog size and per-call token budget tracked in the repo's README and docs/BENCH

Deploys versioned agents, skills, rules, and governed workflows into the native project paths used by Claude Code, Codex, Copilot, Cursor, and other coding agents.

A TypeScript framework for building MCP servers elegantly

RAG based LLM chatbot for dbt projects

Multi-provider AI agent framework that unifies 12+ LLM providers (OpenAI, Google, Anthropic, AWS, Azure, Groq, etc.) with workflow orchestration. Production-grade platform for building LLM applications with streaming, tool calling, caching, and enterprise features. Battle-tested at 15M+ requests/month

Visual prompt builder: decompose prompts into semantic blocks and compile them into Claude-optimized XML

Open-source platform to build, run and govern AI agents in production on ADK and LiteLLM, with RAG, Text-to-SQL, MCP tools, token quotas and per-session audit

Designed with security, scalability, and extensibility at its foundation, AgentUp streamlines development through a configuration-driven architecture and rich plugin ecosystem github

Git-backed platform for skills, tools, and context for AI agents, available to Gemini CLI through a remote OAuth MCP server

MCP SDK for Rust

Open-source Infrastructure for MCP and x402

Self-hosted mission control for AI agents: manage agents, experiments, workflows, crews, skills and more

Self-hosted, multi-frontend agentic harness that runs one persistent agent across Telegram, Discord, Teams, and the terminal. Pluggable backends (Claude, OpenAI Agents, Codex, Kilo, OpenCode), full MCP tool access, and always-on background agents — Goals, Heartbeat, and Dream — backed by a long-term memory palace. MIT, TypeScript

Python framework with 122+ executable AI agent skills, YAML capability contracts, DAG scheduler, MCP server, and adapters for LangChain, CrewAI, and Semantic Kernel

Persistent cross-session memory MCP server for 13+ AI coding tools. SQLite-backed state survives context resets and keeps Claude Code, Cursor, Gemini CLI, Codex, Windsurf, Amp, Kiro, VS Code Copilot, Trae, and more in sync. Install: npm install -g @egchq/egc

Anthropic's Model Context Protocol implementation for Oat++

Integration layer that gives agents per-user OAuth and scoped tool access to 600+ connectors and 35,000+ tools, callable from LangChain, OpenAI Agents SDK, Claude, Hermes or any MCP client.

Give your Gemini CLI agent a wallet, email address, phone number, and 100+ paid MCP tools (web search, image gen, SMS, voice, LLM gateway). Self-register with gemini extensions install https://github.com/atxp-dev/atxp — no human login required, $5 free credits included

CLI tool that transforms any OpenAPI spec into 9 agent interface formats (MCP server, AGENTS.md, CLAUDE.md, .cursorrules, Skills, llms.txt, GEMINI.md, A2A Card, CLI) with a single command. Tiered generation strategies for small to large APIs

A browser engine built for AI agents that compiles HTML into a Semantic Object Model (SOM), providing 10x token compression vs raw HTML. V8 JS rendering, CDP compatibility, authenticated browsing, MCP server github | docs

Read an Obsidian vault directly from disk with compact payloads, no Local REST API plugin required

Type-safe, observable TypeScript AI agent framework on Effect-TS; MCP-native, A2A multi-agent, 6 reasoning strategies, runs the same code on local Ollama 4B+ and frontier APIs

A secure, multi-tenant Python MCP server framework built to integrate easily with external services via OAuth 2.1, offering scalable and robust solutions for managing complex AI applications

Local CLI and stdio MCP payment runtime for AI agents on macOS Apple Silicon, with foreground human approval for signing

Reading, review and guarded editing of DipTrace PCB and schematic projects via a Windows bridge, with cross-platform offline XML analysis

Durable AI agent runtime with MCP client/server and A2A support: graph-based workflows, durable execution and multi-agent coordination, authored in Python

Create a new MCP server in TypeScript, batteries included - supports user-defined templates!

Security-first, self-hosted AI agent platform. Each agent runs in a gVisor sandbox with no network, behind a human approval gateway (deny-by-default); per-session message queues are encrypted with SQLCipher, and releases ship cosign signatures, SBOMs, and build-provenance attestations. AGPLv3, written in Go

Tool retrieval for large tool sets: build a graph from OpenAPI/MCP specs and find multi-step workflows via hybrid search. Also an aggregating MCP proxy

MIT-licensed context compactor for coding-agent transcripts (Claude Code, Codex CLI): moves spent tool outputs to a local content-addressed stash the agent can recall by hash, keeps user and assistant messages byte for byte, and writes a JSON report per pass; CLI, MCP server and Claude Code plugin, offline

Lightweight Java library for building MCP servers using annotations. Annotation-driven development with only 12 dependencies, built-in tools (memory, todo, planner, file ops, bash, telemetry), multi-class modules support, and <500ms cold start

Gin-like Go framework for building MCP servers: struct-tag schemas, middleware, tool groups, Gin/OpenAPI/gRPC adapters, async tasks and an inspector UI

Persistent encrypted memory for AI agents with Trust Quotient scoring, cross-agent handover and an immutable audit trail

Open-source DeepSeek Harness plugin and GitHub Action for progressive MCP tool discovery; its published 1,000-tool benchmark reduces the model-facing MCP interface from 1,000 schemas to 2 and the component schema JSON from 647,962B to 1,114B

Open-source agent-to-agent task marketplace where agents delegate tasks, pick up work, and earn credits. REST API, Python SDK, LangChain/CrewAI/MCP integrations. github | website

Runtime trust scanner for agent skills and tool use: a tested pattern library across 100+ attack categories (prompt injection, tool poisoning, MCP attacks, skill compromise; current counts in the README) plus a mechanism layer, shipped as a pip package, GitHub Action, and free web scanner with a published, reproducible precision/recall benchmark

MCP-native task management control plane for AI agent fleets with multi-agent permissions, delegation safety, and full provenance

curl for x402 paid APIs. CLI and library that auto-pays HTTP 402 responses with USDC on Base and Solana, with MCP stdio proxy for AI agents. npx x402-proxy

YAML-first, event-sourced engine for LLM agents and pipelines. Declarative agents, tools, and connectors run as a DAG; every state change is a hash-chained event, enabling deterministic replay and fork-resume of any past run. Human-in-the-loop approval gates with policy and contracts, MCP and HTTP connectors. Distributed as a Python package

Persistent AI memory for coding assistants. Auto-captures decisions, patterns, and context across sessions. VSCode extension + CLI + MCP server. MIT licensed

Backendless, MIT TypeScript SDK for x402 across 28 chains in 10 families (EVM, Solana, TON, Tron, NEAR, Sui, Aptos, Algorand, Stellar, XRPL). No facilitator, no fee — payments settle straight to your wallet, verified locally against your own RPC. (npm)

Middleware framework for AI agents in TypeScript. Express.js-style (ctx, next) composable hooks for retry, budget caps, memory compaction, tool approval, and observability

#️⃣ ☁️ 🏠 🍎 🪟 🐧 - High-level framework for building MCP servers in .NET with OAuth 2.0/OIDC (6 providers), per-tool MFA, OpenTelemetry, and a native client

Lightweight observability and runtime guardrails for AI agents — loop detection, budget enforcement, cost tracking, and deterministic replay. Zero dependencies, LangChain integration