# AgentStack Daily

> AgentStack Daily: curated agentic AI tools and tips grouped by use case. Private preview.

## Multi-agent

- [CrewAI](https://github.com/crewAIInc/crewAI): Define role-based agent crews with tasks and processes; good for quickly orchestrating researcher/writer/reviewer style teams.
- [AutoGen](https://github.com/microsoft/autogen): Microsoft's framework for conversational multi-agent systems; agents message each other and call tools in event-driven loops.
- [LangGraph](https://github.com/langchain-ai/langgraph): Model agent workflows as stateful graphs with checkpoints, human-in-the-loop interrupts and multi-agent supervisors.
- [OpenAI Agents SDK](https://github.com/openai/openai-agents-python): Lightweight Python SDK with agents, handoffs, guardrails and tracing; use handoffs to route between specialist agents.
- [Google Agent Development Kit (ADK)](https://github.com/google/adk-python): Code-first toolkit for building, evaluating and deploying hierarchical multi-agent apps on Gemini or other models.
- [Agent2Agent (A2A) Protocol](https://github.com/a2aproject/A2A): Open protocol for agents built on different frameworks to discover each other and delegate tasks over HTTP.
- [MetaGPT](https://github.com/FoundationAgents/MetaGPT): Assigns software-company roles (PM, architect, engineer) to agents that collaborate via SOPs to produce a project.
- [CAMEL](https://github.com/camel-ai/camel): Research framework for role-playing communicative agents; useful for studying agent societies and generating data.
- [Swarms](https://github.com/kyegomez/swarms): Enterprise-oriented orchestration library with many swarm topologies (sequential, hierarchical, mixture-of-agents).
- [Anthropic: How we built our multi-agent research system](https://www.anthropic.com/engineering/multi-agent-research-system): Engineering write-up on orchestrator/subagent research: parallel subagents, effort scaling and prompt lessons.
- [Anthropic: Building effective agents](https://www.anthropic.com/research/building-effective-agents): Canonical guide to workflow patterns (prompt chaining, routing, orchestrator-workers) before reaching for full autonomy.
- [Mastra](https://github.com/mastra-ai/mastra): TypeScript agent framework with workflows, agent networks, memory and evals built in.
- [Pydantic AI](https://github.com/pydantic/pydantic-ai): Type-safe Python agent framework; structured outputs and dependency injection make multi-step agents testable.
- [Agno](https://github.com/agno-agi/agno): Fast Python framework for agents and agent teams with memory, knowledge and tool integrations.
- [LlamaIndex](https://www.llamaindex.ai): Framework for document agents and RAG.

## MCP

- [Model Context Protocol spec](https://modelcontextprotocol.io): The open standard for connecting LLM apps to tools, resources and prompts; start here for transports and auth.
- [MCP reference servers](https://github.com/modelcontextprotocol/servers): Official collection of reference servers (filesystem, git, fetch, memory) to learn from or run directly.
- [MCP TypeScript SDK](https://github.com/modelcontextprotocol/typescript-sdk): Official SDK for building MCP servers and clients in TypeScript with stdio and streamable HTTP transports.
- [MCP Python SDK](https://github.com/modelcontextprotocol/python-sdk): Official Python SDK including FastMCP-style decorators for quickly exposing tools.
- [MCP Inspector](https://github.com/modelcontextprotocol/inspector): Developer tool to connect to any MCP server and interactively test tools, resources and prompts.
- [FastMCP](https://github.com/jlowin/fastmcp): Pythonic high-level framework for MCP servers and clients with auth, proxying and composition.
- [GitHub MCP Server](https://github.com/github/github-mcp-server): GitHub's official server exposing repos, issues, PRs and Actions to agents.
- [Playwright MCP](https://github.com/microsoft/playwright-mcp): Microsoft's server that lets agents drive a browser via accessibility snapshots instead of screenshots.
- [MCP Registry](https://github.com/modelcontextprotocol/registry): Official community registry for discovering published MCP servers.
- [Smithery](https://smithery.ai): Hub for finding, hosting and installing MCP servers into clients.
- [Context7](https://github.com/upstash/context7): MCP server that injects up-to-date, version-specific library docs into coding agents' context.
- [Cloudflare remote MCP servers](https://developers.cloudflare.com/agents/guides/remote-mcp-server/): Guide to building and deploying remote MCP servers with OAuth on Cloudflare Workers.

## Computer use

- [Anthropic Computer Use](https://docs.anthropic.com/en/docs/agents-and-tools/tool-use/computer-use-tool): Claude tool that operates a desktop via screenshots, mouse and keyboard; docs cover the reference container.
- [OpenAI Computer Use guide](https://platform.openai.com/docs/guides/tools-computer-use): Docs for the computer-use tool in the Responses API, with a loop for executing actions in your environment.
- [Browser Use](https://github.com/browser-use/browser-use): Open-source library that makes websites accessible to agents; popular for browser automation tasks.
- [Stagehand](https://github.com/browserbase/stagehand): Browserbase's framework mixing Playwright code with natural-language act/extract/observe steps.
- [Browserbase](https://www.browserbase.com): Hosted headless browsers for agents with sessions, stealth and live view.
- [Skyvern](https://github.com/Skyvern-AI/skyvern): Automates browser workflows with LLMs and computer vision; robust to layout changes.
- [OpenAdapt](https://github.com/OpenAdaptAI/OpenAdapt): Open-source process automation that learns from recorded human demonstrations.
- [UI-TARS Desktop](https://github.com/bytedance/UI-TARS-desktop): ByteDance's native GUI agent app built on the UI-TARS vision-language model.
- [E2B Desktop Sandbox](https://github.com/e2b-dev/desktop): Secure cloud Linux desktops for running computer-use agents in isolation.
- [Steel](https://github.com/steel-dev/steel-browser): Open-source browser API for AI agents with session management and proxies.
- [OSWorld](https://github.com/xlang-ai/OSWorld): Benchmark of real desktop tasks for evaluating multimodal computer-use agents.
- [MultiOn](https://www.multion.ai): API for agents that act on the web.
- [Exa](https://exa.ai): Web search API built for AI agents.
- [Tavily](https://tavily.com): Search and extract API optimized for LLM agents.
- [Firecrawl](https://firecrawl.dev): Turns websites into LLM-ready markdown/data.
- [E2B](https://e2b.dev): Secure cloud sandboxes for running agent code.

## Coding agents

- [Claude Code](https://github.com/anthropics/claude-code): Anthropic's agentic coding CLI; works in your terminal with subagents, hooks, MCP and skills.
- [OpenAI Codex CLI](https://github.com/openai/codex): Lightweight coding agent that runs locally in your terminal with sandboxed execution.
- [Cursor](https://cursor.com): AI code editor with background/cloud agents, rules and MCP support.
- [Aider](https://aider.chat): Terminal pair-programmer that edits files in your git repo and commits changes.
- [Cline](https://github.com/cline/cline): Autonomous coding agent for VS Code that plans, edits and runs commands with approval.
- [OpenHands](https://github.com/All-Hands-AI/OpenHands): Open platform for software-development agents (formerly OpenDevin) that can run in a sandbox.
- [Gemini CLI](https://github.com/google-gemini/gemini-cli): Google's open-source terminal agent bringing Gemini to your shell with MCP support.
- [GitHub Copilot coding agent](https://docs.github.com/en/copilot/concepts/agents/coding-agent/about-coding-agent): Assign an issue to Copilot and it opens a PR from an Actions-powered environment.
- [Continue](https://github.com/continuedev/continue): Open-source IDE extensions and CLI for custom coding assistants and agents.
- [Roo Code](https://github.com/RooCodeInc/Roo-Code): VS Code agent fork of Cline with custom modes for architect/code/debug roles.
- [SWE-agent](https://github.com/SWE-agent/SWE-agent): Princeton/Stanford agent that turns LMs into software engineers resolving GitHub issues.
- [Goose](https://github.com/block/goose): Block's extensible local agent that automates engineering tasks via MCP extensions.
- [opencode](https://github.com/sst/opencode): Open-source terminal coding agent supporting many model providers.
- [Claude Code best practices](https://www.anthropic.com/engineering/claude-code-best-practices): Anthropic's tips: CLAUDE.md, plan-first, test-driven loops and running multiple instances.
- [Devin](https://devin.ai): Cognition's autonomous AI software engineer.
- [Replit Agent](https://replit.com): Build and deploy apps from prompts in the browser.
- [Bolt.new](https://bolt.new): Prompt-to-full-stack web app builder in the browser.
- [v0](https://v0.app): Vercel's agent for generating UI and full apps.

## Skills & routines

- [Agent Skills (Anthropic)](https://docs.anthropic.com/en/docs/agents-and-tools/agent-skills/overview): Package instructions, scripts and resources as folders that Claude loads on demand.
- [anthropics/skills](https://github.com/anthropics/skills): Public repo of example and document-handling skills you can copy and adapt.
- [Claude Code hooks](https://docs.anthropic.com/en/docs/claude-code/hooks): Run shell commands at lifecycle events (pre/post tool use, stop) for deterministic guardrails.
- [Claude Code subagents](https://docs.anthropic.com/en/docs/claude-code/sub-agents): Define specialized subagents with their own prompts and tool permissions for delegation.
- [Claude Code slash commands](https://docs.anthropic.com/en/docs/claude-code/slash-commands): Save reusable prompts as custom commands checked into your repo.
- [Claude Code GitHub Actions](https://github.com/anthropics/claude-code-action): Run Claude Code in CI to respond to @claude mentions and automate routine PR work.
- [AGENTS.md](https://agents.md): Open, simple format for giving coding agents project instructions across tools.
- [Cursor Rules](https://docs.cursor.com/context/rules): Persistent project/user rules that shape Cursor agent behavior.
- [n8n](https://github.com/n8n-io/n8n): Workflow automation with AI agent nodes; schedule recurring agent routines with hundreds of integrations.
- [Codex AGENTS.md guide](https://github.com/openai/codex/blob/main/docs/getting-started.md): How Codex reads AGENTS.md files to follow repo conventions.
- [Zapier Agents](https://zapier.com/agents): AI agents that act across Zapier's app integrations.
- [Make](https://www.make.com): Visual automation platform with AI agents.
- [Relevance AI](https://www.relevanceai.com): Build and run teams of AI agents for GTM work.
- [Clay](https://www.clay.com): Data enrichment and AI research agents for outbound.
- [11x](https://www.11x.ai): Autonomous AI SDR and phone agents.

## Memory

- [Mem0](https://github.com/mem0ai/mem0): Memory layer that extracts and retrieves user/agent facts across sessions.
- [Letta (MemGPT)](https://github.com/letta-ai/letta): Stateful agents with self-editing tiered memory, from the MemGPT research.
- [Zep / Graphiti](https://github.com/getzep/graphiti): Temporal knowledge graph for agent memory that tracks how facts change over time.
- [LangMem](https://github.com/langchain-ai/langmem): LangChain library for long-term memory extraction and prompt optimization in agents.
- [Cognee](https://github.com/topoteretes/cognee): Builds knowledge graphs plus vectors from your data as agent memory.
- [MCP Memory server](https://github.com/modelcontextprotocol/servers/tree/main/src/memory): Reference knowledge-graph memory server any MCP client can use.
- [Anthropic memory tool](https://docs.anthropic.com/en/docs/agents-and-tools/tool-use/memory-tool): Claude API tool for storing and retrieving information in a file directory across conversations.
- [Anthropic: Effective context engineering](https://www.anthropic.com/engineering/effective-context-engineering-for-ai-agents): Guide to curating context: compaction, structured notes and sub-agent architectures.
- [Chroma](https://github.com/chroma-core/chroma): Open-source embedding database commonly used for agent retrieval memory.

## Evals

- [SWE-bench](https://github.com/SWE-bench/SWE-bench): Benchmark of real GitHub issues for evaluating coding agents.
- [Inspect AI](https://github.com/UKGovernmentBEIS/inspect_ai): UK AISI framework for LLM and agent evaluations with sandboxed tools.
- [OpenAI Evals](https://github.com/openai/evals): Framework and registry for evaluating LLMs and LLM systems.
- [promptfoo](https://github.com/promptfoo/promptfoo): CLI for testing prompts, agents and RAG with assertions and red-teaming.
- [DeepEval](https://github.com/confident-ai/deepeval): Pytest-like LLM evaluation framework with agentic metrics like task completion.
- [Langfuse](https://github.com/langfuse/langfuse): Open-source LLM observability: traces, evals and prompt management for agents.
- [Arize Phoenix](https://github.com/Arize-ai/phoenix): Open-source tracing and evaluation for LLM apps built on OpenTelemetry.
- [tau-bench](https://github.com/sierra-research/tau-bench): Sierra benchmark for tool-agent-user interaction in realistic domains.
- [Terminal-Bench](https://github.com/laude-institute/terminal-bench): Benchmark for agents completing tasks in real terminal environments.
- [Braintrust](https://www.braintrust.dev): Eval and observability platform for iterating on AI products.
- [Ragas](https://github.com/explodinggradients/ragas): Evaluation toolkit for RAG and agent pipelines.
- [Anthropic: Demystifying evals for AI agents](https://www.anthropic.com/engineering/demystifying-evals-for-ai-agents): Practical guidance on designing graders and tasks for agent evaluation.
- [LangSmith](https://www.langchain.com/langsmith): Tracing, evals and monitoring for LLM apps.
- [W&B Weave](https://wandb.ai/site/weave): Weights & Biases toolkit for tracing and evaluating LLM apps.
- [Helicone](https://www.helicone.ai): Open-source LLM gateway and observability.