Agent Squad is a lightweight orchestration framework that selects a specialist based on intent and conversation context, then preserves the relevant history. The project moved from the AWS Labs repository to 2FastLabs, so the new repository is the source to follow.
Key takeaways
- Focuses on routing conversations among specialized agents.
- Provides Python and TypeScript implementations.
- Moved from
awslabs/agent-squadto2FastLabs/agent-squad.
What is Agent Squad?
Python and TypeScript framework for classifying requests, routing them to specialized agents and retaining per-agent conversation context. It fits assistants that need a clear routing layer across several specialist agents without a full workflow engine, especially when both Python and TypeScript are in use.
What can you build with Agent Squad?
- Classify incoming requests and choose an agent.
- Use streaming or non-streaming agent responses.
- Maintain conversation context for routed specialists.
- Integrate built-in or custom agents and storage backends.
These are documented capabilities, not a guarantee that every model, provider or deployment supports the same behavior. Validate the exact SDK version, model features and tool permissions in a disposable environment before moving a workflow into production.
What is a sensible first project?
Create two harmless specialists with clearly separated intents and a fallback. Test ambiguous, adversarial and out-of-scope messages, then inspect whether context is stored under the correct user and selected agent.
Keep the first run narrow and observable: one input, a small tool allowlist, explicit success criteria, a cost ceiling and a human review point before any external write. Save the prompt, model, SDK version, tool arguments and final result so the test can be reproduced.
How does the architecture handle state and tools?
A classifier considers the request and available agent descriptions, the orchestrator calls the selected agent and a storage layer retains conversation context. Custom agents and classifiers can replace packaged implementations.
Treat model output as untrusted input. Validate structured data, set timeouts and iteration limits, make write operations idempotent where possible, and separate read-only discovery from actions that modify files, infrastructure, customer records or messages.
What should you review before deployment?
- Use the new repository and current packages, not stale AWS Labs links.
- Partition conversation storage by user and agent.
- Enforce tool authorization inside each specialist, not only in the router.
Use least-privileged credentials and isolate code execution, browsers and shell tools. Log tool calls without recording secrets, define an emergency stop, and test how the application behaves when the model, a tool or the network returns an error. Human approval should be enforced in application code for high-impact actions rather than requested only in a prompt.
What are the main limitations?
- It is primarily a router, not a general durable workflow engine.
- Classification mistakes can send sensitive context to the wrong specialist.
- The repository move requires checking package and documentation references.
This profile is based on public first-party documentation checked on 2026-10-04; Anavem did not run a comparative benchmark or a production deployment. APIs, package names, licensing boundaries and hosted services can change, so confirm the current documentation before adopting the framework.
Is Agent Squad the right choice?
Choose it when its programming language, orchestration model and operational controls match a concrete workflow. Compare it with one simpler baseline, including a direct model API plus ordinary application code. The useful decision is not which framework has the longest feature list, but which one makes tool permissions, state, failure handling, evaluation and maintenance understandable to your team.