Connapse

io.github.DestrayonPublished by GitHub account Destrayonv0.3.2Updated Nov 3, 2026

Self-hosted knowledge backend for AI agents with hybrid search and MCP tools

VerifiedSTDIODesktop onlyDatabasesFiles & StorageKnowledge & Memory

Overview

AI-generated overview

Self-hosted knowledge backend that gives AI assistants persistent, searchable memory over your documents through hybrid vector and keyword search.

What you can do
  • Search a knowledge base with hybrid vector and keyword search, optionally reranked, returning answers with source citations
  • Upload files and read documents, and let agents build their own research corpus
  • Use code-related and memory-related tools to keep context across sessions
  • Reach the rest of the REST API through generic tools, alongside 21 named tools
  • Organize material into isolated containers, each with its own files and vector index
When to use it
Useful when you want an assistant to remember past research and query your own documents across sessions, and you are willing to run a self-hosted service rather than rely on a hosted knowledge product.
Requirements
Runs as a local Docker deployment (Docker with Compose plugin, port 5001 free) with PostgreSQL/pgvector, MinIO and optional Ollama; an owner account is created in the web UI, and an agent API key is passed in the X-Api-Key header. Embedding providers such as Ollama, OpenAI, Azure OpenAI or Anthropic can be configured.
Before you install
The agent key is a credential sent in the X-Api-Key header; keep it private. Agents can upload files and write to the knowledge base, and generic tools reach the wider API. Connecting cloud sources such as S3, Azure, GitHub, Confluence or Jira sends data to those services, and using a hosted embedding provider sends content to that provider unless Ollama is used.

Installation

In SourceWeft

  1. Open Connapse in the dashboard and add it to a workspace.
  2. Enable the server for the chats that should use its tools.

Desktop only via STDIO. STDIO servers start a local process, so they need the SourceWeft desktop host.

Other MCP clients

Follow the launch instructions in the repository.

README

[Connapse logo]

Stop losing context between AI sessions. Give your agents persistent, searchable memory.

[License: MIT] [.NET] [Build] [Tests] [PRs Welcome] [GitHub Issues] [GitHub Stars] [Connapse MCP server] [Docker]

[Connapse demo — upload a PDF, search with hybrid vector and keyword search, get results with source citations in seconds]

Your AI agents forget everything between sessions. Connapse fixes that.

Every time you start a new conversation, your AI agent starts from zero — no memory of past research, no access to your documents, no accumulated knowledge. Connapse is an open-source knowledge backend that gives agents persistent, searchable memory. Upload documents or point it at your existing Amazon S3 buckets or local filesystems. Agents query and build their own research corpus through the MCP server, the REST API, or the CLI. Container-isolated, hybrid search (vector + keyword), self-hosted and private. Deploy in 60 seconds with Docker. Built on .NET 10.

🤖 AI Agent Integration — Claude queries and builds your knowledge base via MCP

[Claude querying Connapse knowledge base via MCP server — asks about preventing cascading failures in microservices, gets structured answer with circuit breaker pattern details cited from distributed-systems-notes.md]

AI agents query your knowledge base through the MCP server, receiving structured answers with source citations from your documents.

🎛️ Your Knowledge, Your Rules — Runtime configuration without restarting

[Connapse settings panel — switching embedding providers, adjusting chunking parameters, and configuring search settings at runtime without restart]

Switch embedding providers, tune chunking parameters, and configure search — all at runtime, without restarting.

📦 Quick Start

You need Docker with the Compose plugin and port 5001 free.

bash
git clone https://github.com/Destrayon/Connapse.gitcd Connapsecp .env.example .envdocker compose --profile with-ollama up -ddocker compose exec ollama ollama pull nomic-embed-text

Open http://localhost:5001 and create the owner account. The full walk-through, from upload to a cited search result, is the Quickstart.

🤖 Connect an agent

Create an agent key at Agents → API Keys, then add Connapse to your MCP client. For Claude Code:

bash
claude mcp add --transport http connapse http://localhost:5001/mcp --header "X-Api-Key: cnp_your_key"

Agents get 21 named tools, such as search_knowledge, upload_file, get_document, the code_* tools and the memory_* tools, plus 3 generic tools that reach the rest of the API. Other clients are in Connect an MCP client, and every tool is in MCP tools.

🚀 Features

  • 🗂️ Containers — Each project gets its own files, vector index and optional settings, with nothing shared between containers.
  • 🔍 Hybrid search — Vector and keyword search fused together, with optional cross-encoder reranking.
  • 🧠 Your choice of models — Ollama, OpenAI, Azure OpenAI and Anthropic, changed at runtime without a restart.
  • 🔌 Index what you already have — Uploads, local folders and SFTP, Amazon S3, Azure Blob Storage, GitHub, Confluence and Jira, and your own systems through a custom source server. Files stay where they are.
  • 🔐 Per-user permissions — Search results from AWS, Azure, GitHub, Atlassian and custom sources are trimmed to what each person can open in the source system, and fail closed.
  • 🤖 Every surface — Web UI, REST API, MCP server and the connapse CLI.
  • 🐳 Self-hosted — Docker Compose with PostgreSQL and pgvector, MinIO and optional Ollama. With Ollama, nothing leaves your machine.

📚 Documentation

The full documentation is in docs/.

  • Get started — run Connapse and search your first document
  • Self-host — install, production, backups, upgrades and troubleshooting
  • Connect your data — sources from S3, Azure, GitHub, Atlassian, folders and your own systems
  • Guides — users, tokens, agents, the API and search tuning
  • Concepts — how Connapse, search, ingestion and permissions work
  • Reference — REST API, MCP tools, configuration, roles and scopes, file types and error types
  • Security policy

💬 Support

🤝 Contributing

Contributions are welcome. Start with CONTRIBUTING.md and Development, and look for issues labeled good-first-issue.

📄 License

MIT.

Source: README.md at commit 8dd7a27

Tools

0
SourceWeft has not started this package in its sandbox yet, so its tools are not known.

Version history

1
  1. v0.3.2LatestSep 16, 2026

Score

50% of the applicable checks met

How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.

  • Reached in the last 30 days25 pts

    Keep the endpoint answering the MCP handshake; SourceWeft probes it every 30 days.

  • Auth declared10 pts

    Declare the credential headers in the registry entry, or keep the endpoint answering the probe so its auth can be classified.

  • Tools listed15 pts

    Answer tools/list without a credential, or ship a package that starts in a sandbox without one.

  • DescriptionRequired15 pts
  • Logo10 pts
  • Documentation15 pts
  • License10 pts