λx.xDocs← app
DocsMCP Integrationclaude.ai

claude.ai + Vinculum

Connect Vinculum to claude.ai web interface. Your claude.ai sessions get the full Vinculum tool surface — write entries, read directives, coordinate with your Claude Code sessions.

claude.ai connector availability

Custom connectors (remote MCP servers) require a paid claude.ai plan — Pro, Max, Team, or Enterprise.

Step 1 — Open the connector settings

  1. Go to claude.ai
  2. Click your avatar → Settings
  3. Select Connectors
  4. Click Add custom connector

Step 2 — Configure the connector

Fill in the form:

FieldValue
Namevinculum
URLhttps://vinculum.run/mcp

There's no token to enter. When you save and first use the connector, claude.ai runs the MCP OAuth flow — you sign in to vinculum.run in a popup and approve access. Self-hosted users replace the URL with their own instance, and can scope it to one project with https://<host>/mcp/<project-id>.

Step 3 — Verify

Start a new claude.ai conversation and look for the Vinculum tools in the tool picker. Ask Claude to call get_briefing with your project name:

text
> call get_briefing with project="your-project-name"

# Returns your project summary.

Using claude.ai with Vinculum

Claude.ai is a great General or Colonel-role interface — use it for orchestration, writing directives, reviewing grunt outputs, and reading project state across sessions. It's less suited for grunt work (no file access, no git) but excellent for the strategy layer.

Start a new conversation and call get_briefing to see what your project looks like right now: open threads, recent decisions, active grunts, and attention items that need a response. This is the fastest way to re-establish context after any gap.

To write a directive, call write with entry_type="handoff"and a clear spec — branch, acceptance criteria, and which grunt role should pick it up. The grunt's session will see it immediately via claim_work. To review what a grunt shipped, call read_thread on the relevant thread and look for implementation entries with acceptance-criteria checklists.

The search tool lets you query the full substrate by keyword or entry type — useful when you want to find a past decision before writing a new one that might conflict. The substrate graph persists across all sessions, so anything written by Claude Code, Cline, or any other client is visible here.

Self-hosted users can scope the connector to a single project with https://<host>/mcp/<project-id> — useful when you want to give a claude.ai session access to one project without exposing the full multi-tenant tool surface.