What is Orcha

Orcha stores durable knowledge once and makes it available to any authorized AI tool. Files, structured databases, and read-only connected sources live in one workspace, and agents reach them through the web app, a REST API, or the Model Context Protocol (MCP), with permissions, citations, and provenance.

What Orcha stores

• Files: Markdown, JSON, and YAML files in a nested folder tree. See Files and folders.
• Structured databases: Typed schemas with relations, saved views, and versioned records. See Databases.
• Connected sources: Read-only documents synced from GitHub, Google Workspace, Microsoft 365, Dropbox, Notion, Atlassian, Linear, Zoom, and Granola. You choose what gets indexed per repository, folder, or path, and citations pin to the exact upstream version. See Connected sources.

How agents reach it

• REST API: Token-authenticated endpoints under /api/v1. See the API reference.
• MCP server: An MCP server over stateless Streamable HTTP at /mcp, compatible with Claude Code, Claude Desktop, Cursor, and any MCP client. See the MCP reference.
• Hybrid retrieval: pgvector semantic search fused with PostgreSQL full-text search, returning line-anchored results with citations. See Retrieval.
• Filesystem-shaped access: A read-only virtual filesystem over the workspace, browsable with ls, tree, grep, and cat through the orcha CLI, REST, or MCP. See Browse context like a filesystem.

Where to start

1. Quickstart: create a workspace, mint a token, and make your first API call.
2. Connect an agent: Claude Code, Claude Desktop, or Cursor.
3. Connect a source and choose what gets indexed, or import from Obsidian or Notion to bring existing knowledge with hierarchy, relations, attachments, and links preserved.