Team Relay / AI Integration

Your Obsidian vault,
accessible to AI

MCP server for Team Relay. Read, search, and write notes from Claude Code, Codex CLI, Cursor, or any MCP-compatible tool.

AI agents can't reach your knowledge base

Your knowledge base is right there. Your AI just can't access it.

Specs, architecture docs, meeting notes, project plans — all in Obsidian.

Your AI coding agent has no idea they exist.

You copy-paste context into prompts. Manually. Every time.

The agent hallucinates what it can't see.

MCP —a standard protocol for AI-to-vault access

Model Context Protocol (MCP) is an open standard for connecting AI tools to external data. Team Relay implements MCP so any compatible agent can read and write your vault through a single integration. No custom plugins per tool. No file system hacks.

Read

Notes, folders, and search results

Write

New docs or update existing ones

List

Files and folder structure

Search

Across the entire vault

How it works

1

Install

$ pip install evc-team-relay-mcp

Also available via Docker or as an OpenClaw skill.

2

Configure

{
  "mcpServers": {
    "team-relay": {
      "command": "uvx",
      "args": ["evc-team-relay-mcp"]
    }
  }
}

Add the MCP server to your AI tool config.

3

Use

"Read the project spec and suggest architecture"
"Write meeting notes to /meetings/today.md"
"Search vault for auth implementation details"

Your AI agent now has vault access.

For autonomous AI agents on OpenClaw

If you run OpenClaw —the self-hosted AI agent platform —Team Relay is available as a ready-made skill. Install in one line:

$ openclaw skill install venturecrew/evc-team-relay
agent.yaml
skills:
  - name: venturecrew/evc-team-relay
    credentials:
      relay_url: https://cp.tr.entire.vc
      api_key: ${{ secrets.RELAY_API_KEY }}
Daily standup bot

Agent reads yesterday's notes, generates a standup summary, writes it to the team log.

Docs-aware coding assistant

Agent reads project specs from the vault before writing any code. Always up to date.

Meeting note processor

Agent takes raw meeting notes, extracts action items, updates relevant project pages.

Use cases

Feed project context to coding agents

Point Claude Code or Codex at your project specs. The agent reads the latest version from your vault —not a stale paste from three weeks ago.

Before (manual)
// Copy-paste 200 lines of spec into prompt
// Spec changes next day — agent doesn't know
// Hallucinated requirements everywhere
After (MCP)
"Read the spec at /projects/acme/spec.md
 and implement the auth module"
// Agent reads live spec, always current

Generate documentation automatically

CI pipeline writes API reference to your vault via MCP. Deploy triggers doc generation.

# In CI pipeline
evc-mcp write \
  --path "/docs/api/v2-reference.md" \
  --content "$(generate-api-docs)"

Build AI workflows on team knowledge

Search your vault programmatically, pull context into custom AI pipelines. Your team's knowledge becomes structured input for any automation.

Keep client-facing docs in sync

AI reads your internal specs, rewrites them for an external audience, and publishes. Internal and public docs always aligned.

Onboard new team members

"Read the onboarding folder and answer my questions." New hires get answers grounded in your actual docs, not outdated wikis.

Configuration

Add Team Relay MCP to your tool of choice. Pick your setup:

claude_desktop_config.json
{
  "mcpServers": {
    "team-relay": {
      "command": "uvx",
      "args": ["evc-team-relay-mcp"],
      "env": {
        "RELAY_CP_URL": "https://cp.tr.entire.vc",
        "RELAY_API_KEY": "your-api-key"
      }
    }
  }
}

How it compares

Manual copy-paste RAG pipeline File system access Team Relay MCP
Setup time None Hours-days Minutes Minutes
Always current Depends on indexing
Write back to vault
Works with any MCP tool
Access control N/A Custom OS-level API key + per-folder
Team collaboration
Search grep/find Full-text

Manual copy-paste —no setup, but stale context and no write-back. Does not scale.

RAG pipeline —powerful search, but requires embedding infrastructure, indexing lag, and no writes.

File system access —works locally, but no team sync, no remote access, no access control.

Get started

Fastest

Hosted

  1. 1 pip install evc-team-relay-mcp
  2. 2 Sign up at entire.vc
  3. 3 Add config to your AI tool
  4. Done
Get started

Self-hosted

  1. 1 Deploy Team Relay on your VPS
  2. 2 pip install evc-team-relay-mcp
  3. 3 Point config to your server URL
  4. Done
View on GitHub

OpenClaw

  1. 1 Install the skill
  2. 2 Add credentials to agent config
  3. 3 Agent has vault access
  4. Done
OpenClaw Skill

Give your AI the context it needs

Stop copy-pasting. Start building with real knowledge.