Production AI Research Agent
The ResearchAgent is a working AI research agent: it searches real sources, synthesizes what it finds, cites it, and exports a report — with the full production layer of memory, tracing, guardrails and evals around it, so the output is auditable, not a black box.
Grounding
A retrieval-grounding score flags claims the sources don't actually support.
Observability
Every search and synthesis step is in the step tree with its cost — a long run is inspectable and its spend is known.
Scratchpad memory
Findings can be parked in the scratchpad tier for long autonomous runs that outgrow the context window.
Evals
Seeded cases score citation quality and grounding so the agent doesn't drift toward confident-but-unsourced answers.
agentfast new research --sdk vanilla
docker compose upThe ResearchAgent ships on the vanilla loop — agentFast's own tool-use loop with the entire production layer around it. See also the support agent and coding agent, or the tools that power the search.
agentFast is the production layer — memory, 50 tools, observability, evals, guardrails, human-in-the-loop — for LangGraph, CrewAI, the Claude & OpenAI Agent SDKs & Vanilla. Own it for life.
Get agentFast — from $89