Agent Integration¶
Docuflow integrates with all major MCP-compatible AI agents. docuflow init handles registration automatically.
Claude Desktop¶
docuflow init adds an entry to ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or the equivalent on Windows/Linux.
Manual registration (if needed):
Restart Claude Desktop after editing the config.
Using Docuflow in Claude:
docuflow init generates a CLAUDE.md file at your project root. Claude reads this automatically and knows which tools to use and when. You can also prompt directly:
VS Code + GitHub Copilot¶
docuflow init creates .mcp.json in your project root:
Copilot CLI reads this automatically in the project directory. Tools are available with the docuflow- prefix:
Cursor¶
docuflow init creates .cursor/mcp.json:
In Cursor's AI chat, reference tools by name or describe what you want:
Checking registration¶
The doctor command reports:
- Installed package versions
- Which agents have Docuflow registered
- Whether
docuflow-studiobinary is resolvable - Wiki health for any discovered projects
Using the MCP tools¶
Once registered, you interact with Docuflow naturally in your AI agent. Common patterns:
Before coding¶
Before I write this feature, check the wiki with query_wiki: "What are the existing patterns for database access?"
After adding a document¶
I just dropped architecture-notes.md in .docuflow/sources/. Ingest it and tell me what new entities were discovered.
Checking health¶
Getting suggestions¶
Waymark security integration (optional)¶
Waymark is a security policy enforcement layer for MCP-based AI agents. When Waymark is configured alongside Docuflow:
- File and shell operations go through Waymark for policy enforcement
- Sensitive paths (
.env, secrets) are blocked automatically - Dangerous commands (
rm -rf,sudo) are blocked - Sensitive operations require human approval via the Waymark dashboard
- Full audit trail of every AI action
See the Waymark documentation for setup.