MCP MCP server · io.github.BrowseAI-HQ
io.github.BrowseAI-HQ/browseai-dev
Evidence-backed web research for AI agents with citations and confidence scores.
- host stars
- 20
- 30-day movement
- starts with the next reading
- Related entries
- 61
- Connections
- 13
npmlocalstdio
BrowseAI Dev is an MCP server (stdio transport) exposing tools for evidence-backed web research, including search, open, extract, answer, compare, clarity, and persistent session tools (create, ask, recall, share, knowledge, fork) plus feedback. Answers pass through a verification pipeline that attaches citations and evidence-based confidence scores, with depth modes, query planning, and retry with backoff.
You want AI agents to produce web research answers where every claim is tied to a real source with a confidence score.
Use it to
- Run cited web searches from an MCP client
- Maintain persistent research sessions across queries
- Decompose complex queries into planned sub-queries
- Submit good/bad/wrong feedback on results
- Stream research progress via SSE
For Developers building AI agents that need cited web research
- Runs
- installed — on your machine
- Transport
- stdio
- Tools
- 13
- Packaged as
- npm:browseai-dev
- Install
- npx -y browseai-dev
- Needs
- 1 environment variable, including secrets
- Version
- 0.4.0
- Last release
- 2026-09-07
topicsmcpweb-researchcitationsverificationsearchsessions