DocBrain MCP Server

DocBrain ingests knowledge from every tool your team uses, answers questions with source attribution and confidence scoring, and autonomously identifies documentation gaps — turning every unanswered question into a documented solution.

Website • Quickstart • Features • Architecture • Docs • Community • Contributing

Project Status: DocBrain is currently distributed as pre-built Docker images and deployment artifacts (Helm charts, configuration, documentation). Source code is not yet published. CI/CD pipelines, build-from-source instructions, and automated test suites will be added when the source is released. Contributions are currently welcome for documentation, configuration, and bug reports against the published artifacts.

Overview

DocBrain is a RAG-based documentation intelligence platform built in Rust. It connects to 13+ knowledge sources (Confluence, Slack, GitHub, Jira, PagerDuty, Zendesk, Microsoft Teams, and more), provides confidence-scored answers with source attribution, and runs an autonomous Autopilot that detects documentation gaps and drafts missing content.

Unlike static search tools, DocBrain maintains a multi-tier memory system that compounds over time — every question, answer, and feedback signal makes the next response better.

See It In Action

Key Features

• 13+ Knowledge Sources — Confluence, Slack, Microsoft Teams, GitHub PRs, GitLab MRs, Jira, PagerDuty, OpsGenie, Zendesk, Intercom, local Markdown, and more
• Confidence-Scored Answers — Zero-guess policy: high confidence returns sourced answers, low confidence asks clarifying questions instead of hallucinating
• Documentation Autopilot — Autonomously clusters unanswered questions, detects gaps, and drafts missing documentation using your org's existing voice
• 4-Tier Memory System — Working, episodic, semantic, and procedural memory that compounds with every interaction
• Document Health Scores — 5-signal freshness scoring (time decay, engagement, content currency, link health, contradiction detection) with proactive staleness alerts
• Cross-Document Reference Graph — Automatically extracts and links references across documents (GitHub PRs, GitLab MRs, Jira tickets, Confluence pages) for richer context during retrieval
• Real-Time Capture — /docbrain capture in Slack threads, @docbrain capture on GitHub PRs and GitLab MRs for instant knowledge indexing
• Intent-Adaptive Responses — Classifies queries (find, how-to, troubleshoot, who-owns, status, explain) and adapts response format accordingly
• Image Intelligence — Vision-capable LLM extraction of architecture diagrams, flowcharts, and screenshots during ingestion
• Multi-Team Space Isolation — Soft boost, per-request filters, and API-key-level hard restrictions for multi-tenant deployments
• Multiple LLM Providers — Anthropic, OpenAI, AWS Bedrock, Ollama (fully local), Google Gemini, Vertex AI, DeepSeek, Groq, Mistral, xAI, Azure OpenAI, OpenRouter, Together AI, and Cohere

Intelligence Layer

DocBrain's intelligence layer goes beyond retrieval with five systems that make it proactive, self-improving, and organizationally aware:

Knowledge Graph

BFS/DFS traversal over your entity graph surfaces structural knowledge: "What depends on the auth service?", "Who are the experts on Kubernetes?", "What's the blast radius if Redis goes down?" Graph traversal answers questions that no amount of vector similarity can.

API: GET /api/v1/graph/entity/:name, `GET /api/v1/graph/blast