Which framework should I use to build my own agent?

+ Code-as-action agents in about a thousand lines of core code that work with any Hub model.
− Running model-written Python needs sandboxing, which is left mostly to you.

+ Unified provider interface and React hooks make streaming chat UIs trivial.
− Major-version upgrades have required significant rewrites.

+ MCP-first design means any MCP server becomes a Goose capability with no glue code.
− Less tuned for coding than Claude Code, so results vary widely by model.

+ Explicit state graphs with checkpointing make long-running agents debuggable and resumable.
− Steep learning curve and more boilerplate than simpler agent SDKs.

+ Largest integration catalogue of any framework, so almost any provider is one import away.
− Frequent breaking API changes and heavy abstractions make simple things feel over-engineered.

+ Inherits Claude Code's file, shell, and subagent tooling, so agents are production-grade on day one.
− Tied to Claude models and the Claude Code runtime rather than a provider-agnostic design.

+ Best-in-class document loaders, indexes, and LlamaParse for messy PDFs.
− Agent features are secondary to retrieval and the package surface is sprawling.

+ Role, goal, and task abstractions make multi-agent setups readable in a few lines.
− Agents can loop or burn tokens unpredictably with little low-level control.

+ Templates for sales and support agents work with Gmail, Slack, and HubSpot out of the box.
− Credit-based pricing is hard to forecast for always-on agents.

+ Python-native components can be edited inline, so flows are not limited to built-in nodes.
− Several critical security vulnerabilities in 2025 made exposed instances risky.

+ AutoGen Studio offers a no-code UI for prototyping agent teams.
− The v0.4 rewrite broke compatibility and the AG2 fork split the community.

+ Covers prompt management, knowledge bases, and deployable APIs in one self-hostable app.
− Complex workflows get unwieldy in the visual canvas and version control is weak.

+ Fastest way to prototype a LangChain chatbot and expose it as an API or embed widget.
− Inherits LangChain's churn and breaks when node versions drift.

+ First-class deployment to Vertex AI Agent Engine and support for the A2A protocol.
− Documentation and examples skew heavily to Google Cloud.

+ Batteries-included TypeScript stack with a local dev playground for inspecting agent runs.
− Newer and smaller than LangChain, with APIs still shifting.

+ Handoffs and built-in tracing are simple to reason about and integrate with the OpenAI dashboard.
− Tuned for OpenAI models; other providers work but lose features.

+ Validated, typed outputs and a FastAPI-like developer experience with minimal magic.
− Younger ecosystem with fewer integrations than LangChain.