connerlambden/bgpt-mcp — Scientific paper search API for AI agents: REST, Python, OpenAPI, and MCP with structured full-text…

Description and use

**Brief Summary**
connerlambden/bgpt‑mcp is an open‑source REST/Python API that implements the Model Context Protocol (MCP), letting AI agents retrieve scientific papers with structured full‑text evidence. It provides a standards‑based, query‑able backend (OpenAPI spec, SDK, CLI) that can be deployed as a self‑hosted MCP server or consumed directly from Python.

**Value**

• **Standardised bridge** – By speaking MCP, the service lets any LLM‑based assistant access vetted scientific literature without custom scraping or ad‑hoc parsers.
• **Structured evidence** – Returns not just raw PDFs but parsed sections, citations, and metadata, enabling RAG pipelines to cite sources accurately.
• **Multi‑language support** – Although the core is JavaScript, there are first‑class Python bindings and a full OpenAPI contract, making integration trivial for both backend services and AI‑first applications.

**Practical Adoption Path**

1. **Prototype** – Pull the Docker image or run npm install to spin up a local MCP server; use the provided Python SDK to issue a paper‑search request and inspect the JSON evidence payload.
2. **Integrate** – Replace any existing “search‑and‑scrape” component in your agent with a thin wrapper around the SDK/CLI; the API’s /search and /evidence endpoints map cleanly to typical RAG steps (retrieve → augment → generate).
3. **Scale** – Deploy the server behind a load balancer or as a Kubernetes micro‑service; configure the built‑in PostgreSQL (or plug in your own DB) for caching and citation indexing.
4. **Publish** – Offer the MCP endpoint to third‑party agents or partner teams, leveraging the OpenAPI spec for automatic client generation in other languages.

**Production Readiness**

• **Activity & Community** – 36 ★, 9 forks, recent commits (last update 2026‑07‑06), and a healthy set of 20 topics indicate an active maintainer base.
• **Architecture** – Clean separation of API, SDK, and CLI; OpenAPI definition enables contract testing and CI pipelines.
• **Deployment Maturity** – Docker images and Helm charts are provided, and the JavaScript core can run in any Node.js‑compatible environment, simplifying cloud or on‑prem deployment.
• **Risk Checklist** – No glaring licensing or security red flags yet, but a final audit of the MIT/Apache license terms, dependency vulnerabilities, and maintainer responsiveness is advisable before a full production rollout.

Overall, connerlambden/bgpt‑mcp is a solid, production‑grade building block for any organization that wants to empower LLM agents with reliable, citation‑ready scientific knowledge.

Problem

connerlambden/bgpt-mcp helps connect AI assistants to real tools and data through a standard protocol.

Outcome

High for an OSS candidate: recent activity, adoption, and ecosystem signals are strong enough for a serious pilot.

Workflow

• connect AI agents to tools
• ship Model Context Protocol servers
• standardize integrations

Evidence

Scientific paper search API for AI agents: REST, Python, OpenAPI, and MCP with structured full-text evidence.

Source