Biolab
Auditable API and MCP infrastructure for agent-facing retrieval
A Python and Go MCP service for agent-facing scientific retrieval that preserves verbatim source payloads and hash-chains each retrieval to the preceding audit record.
Python
Go
FastMCP
SQLite
Turso
Docker
GitHub Actions
Hugging Face Spaces
PyPIAI agents can query scientific databases without leaving enough evidence to reconstruct the exact source response, query, and retrieval context behind a later claim.
Five MCP tools expose PubMed, Europe PMC, ClinicalTrials.gov, and bioRxiv/medRxiv through one agent-facing interface.
Verbatim upstream responses are stored alongside normalized audit fields for reproducibility.
Each retrieval hash includes the previous record hash; migration backfills existing histories and verifies every link in sequence.
API keys and rate limits protect the public service boundary while keeping local MCP workflows straightforward.
Python and Go tests verify deterministic hashing, concurrent writer serialization, migration backfills, and detection of modified or missing interior records. The Python SDK is published to PyPI and the MCP server is listed in the official registry.