Archonics performs context engineering audits on production AI agent systems. We examine discoverability, system prompts, tool definitions, context packing, and evaluation infrastructure — the five dimensions that decide whether an agent built in a demo can be found, called, and survive real traffic.
Before the four dimensions below matter, a service has to be discoverable to the agents that would invoke it. We examine presence across MCP Registry, x402scan, CDP Bazaar, agentic.market and the GitHub topic graph; correctness of the discovery JSONs against each indexer's parser; brand-impersonation risk; and the agent-readable surfaces (llms.txt, /agents.json) a calling agent reads on landing.
GPT-Researcher has twenty-six thousand stars and sixty-nine releases. It is architecturally sound and actively maintained. It also has nineteen findings in a rigorous audit — two of them critical.
Read the full report. It's how we'd write one for your team, at the same depth, applied to your system prompt, your tool definitions, your context, your evals. No sales call, no gated form, no marketing warmth — just the audit.
Read the full audit (PDF) →