Frequently asked questions
Everything teams ask before adopting FastAgentic. Still stuck? Get in touch.
Basics
- FastAgentic is the deployment layer for agentic applications — an open-source (MIT) Python runtime built on FastAPI. It wraps agents written in PydanticAI, LangGraph, CrewAI, or LangChain and exposes each one as a REST endpoint, an MCP tool, and an A2A skill from a single decorator, with auth, durable checkpoints, cost tracking, and observability built in.
- Yes. FastAgentic is MIT-licensed and self-hosted. There is no proprietary control plane and no cloud lock-in — you run it wherever you run Python.
- Nothing. FastAgentic is free and open source. Neul Labs, the team behind it, offers optional paid consulting for teams that want help shipping agents to production.
- Python 3.10 or newer. FastAgentic is built on FastAPI, so it runs anywhere FastAPI runs — a container, a VM, or a serverless Python host.
What is FastAgentic?
Is FastAgentic open source?
What does it cost?
What are the requirements?
Frameworks & protocols
- PydanticAI, LangGraph, CrewAI, and LangChain have first-class adapters. Anything else can be wrapped through the Runnable interface, and run_opaque() can lift an existing agent object without touching its code.
- REST is the standard HTTP API surface (with streaming). MCP (Model Context Protocol) exposes your agent as a tool that assistants like Claude and Cursor can call. A2A (Agent-to-Agent) advertises skills so other agents can collaborate with yours. FastAgentic generates all three from one endpoint definition with shared schemas.
- No. run_opaque() wraps an existing agent unchanged, snapshotting inputs and outputs to make the run resumable. You keep your current code and gain deployment features on top of it.
- Yes. Because deployment, auth, durability, and observability live in FastAgentic rather than your framework, you can swap the underlying agent framework without rewriting any of that surrounding infrastructure.
Which agent frameworks does FastAgentic support?
What are REST, MCP, and A2A?
Do I have to rewrite my existing agents?
Can I switch frameworks later?
Production
- StepTracker checkpoints every workflow step to Redis, Postgres, or S3. If a process restarts mid-run, execution resumes at the step level — no external orchestrator like Celery is required.
- FastAgentic tracks token usage and pricing per run, per user, and per tenant, and enforces hard budget cut-offs before spend happens. RBAC, scope checks, and PII masking round out the governance layer.
- OpenTelemetry spans across every step, structured JSON logging with full request context, and native integrations for Langfuse, Portkey, and Datadog.
- LangServe is tied to LangChain. FastAgentic is framework-agnostic, adds MCP and A2A protocol surfaces, ships durable step-level checkpoints, and includes cost, policy, and audit governance that LangServe does not. See the LangServe comparison for a full breakdown.
How does FastAgentic handle long-running or crashing runs?
How does cost control work?
What observability does it provide?
How is FastAgentic different from LangServe?
Need FastAPI, LangGraph, or agent platform expertise?
Neul Labs — the team behind FastAgentic — takes on a limited number of consulting engagements each quarter. We help teams ship agents to production, fix broken LangGraph pipelines, and design governance for multi-tenant LLM platforms.