Deterministic, verifiable data for agents.
One hosted MCP server and one CLI over the primary biological databases — the same correct result every run, with a manifest you can re-run and verify(). No portal to scrape.
Public reads need no key — add your pk_live_ key for an account. Any MCP client, the full config, and pinakes mcp install are in the docs.
Deterministic
Same query, same bytes, every run — snapshot-pinned and provable with a manifest.
Complete
Full retrieval, or it fails loudly. Never a silent truncation — the bug that put one outbreak’s origin anywhere from 1922 to 2014.
Agent-native
A typed MCP tool, not a website to click through.
See it run
Prefer the shell? Install the CLI, then watch determinism end to end — retrieve records, then verify the manifest reproduces them byte-for-byte.
Also pip install pinakes · brew install pinakes · conda. The command is always pinakes. Pre-launch: mcp.pinakes.sh, the packages, and the installer publish at launch — the commands above are the real, final shapes. See GitHub for current availability.
Biological databases for AI agents
Biology’s primary databases — plus arXiv literature — each served the same way, snapshot-pinned and provable with verify(). Ten ship in the first wave; 64 in all.
A preview of the intended catalog surface — connectors are in active development, not a live count.
Get a key
Free while Pinakes is in preview — the full catalog, reproducible retrieval, and verify(), all at the base rate limits. No tiers, no card.
Public reads need no key — add a pk_live_ key for an account. Everyone shares the same base rate limits while we’re in preview.
Biology today. The same deterministic engine, any corpus — named for the Pínakes, the first catalogue of the Library of Alexandria.