BotsHub Agent¶
An autonomous agent for botshub.com that finds which AI agents are getting popular and builds a directory listing for each one, using only facts it actually fetched.
It is built on AAA, the Autonomous Agent Architecture library
(pip install autonomous-agent-architecture), and this site doubles as a worked, real-world example of using AAA:
every pipeline stage maps to an AAA algorithm, and the lessons page records what we learned running it against live services.
-
:material-magnify: Finds what is trending
GitHub, Hacker News and SearXNG signals, stored with dates so we can compute star velocity, not just popularity. Discovery →
-
:material-shield-check: Never invents
Every claim in a listing carries a verbatim quote from a page we downloaded. A validator rejects anything it cannot find. Grounding →
-
:material-sqlite: SQLite everywhere
Business data in one SQLite file, AAA's event log in another. One machine, crash-safe, no servers to run. Data →
-
:material-lightning-bolt: Built on AAA
WorkflowGraph,GenerateTestRepair,GeneratorCritic,SelfConsistencyandToolLoop, each used where it fits. Using AAA →
What it produces¶
For each agent it finds (like aiagentslist.com-style listings): name, website, overview, features, use cases, integrations, pricing, screenshots/videos and specs. Output is JSON per agent plus an index, ready for botshub.com to consume. See the export format.
flowchart LR
S[GitHub · HN · SearXNG] --> D[Discovery<br/>score + dedupe + classify]
D --> C[(candidates<br/>SQLite)]
C --> E[Enrichment<br/>fetch → extract → verify]
E --> G{Publish gate}
G -->|passes| P[(listings + JSON)]
G -->|fails| H[held, with reasons]
Try it¶
git clone https://github.com/innerkorehq/botshub-agent && cd botshub-agent
uv sync --extra docs
cp .env.example .env # add your LLM gateway key
uv run botshub discover && uv run botshub enrich --limit 3
Details in the quickstart. The project is alpha; see status and known gaps for what is and is not verified.
About AAA
AAA is an async, event-driven Python library for the non-model parts of an agent system: loops, plans, routing, retries, budgets, approvals, persistence and recovery. Read its documentation or star it on GitHub.