WrenAI MCP

Connect AI agents to WrenAI — query your data, generate SQL, build charts, and get business insights in natural language, straight from Claude or ChatGPT.

AI client (Claude / ChatGPT)
   │  OAuth
   ▼
https://cloud.getwren.ai/api/mcp          ← one endpoint for your whole org
   ├── list_projects            (discover which projects you can query)
   ├── get_project_metadata     (table schema for a project)
   ├── ask                      (end-to-end natural-language Q&A)
   ├── generate_sql / run_sql
   ├── generate_chart           (Vega spec)
   └── generate_summary

New model (July 2026): MCP is now a single organization-level endpoint with OAuth login — no API key and no local proxy. After you connect, your agent calls list_projects and then works against any project your org has exposed. The old per-project URL (…/api/mcp/{token}) is deprecated — see Migrating.

---

How it works

1. An org admin enables MCP for the organization and picks which projects are exposed.
2. You add one URL — https://cloud.getwren.ai/api/mcp — as an MCP connector in your AI client and log in with OAuth (your WrenAI account).
3. Your agent calls list_projects to see the projects you can reach, then passes a projectId to the other tools.
4. Data security (row/column-level) is applied per user, server-side — you only see what your WrenAI account is allowed to see.

Requirements

• A WrenAI Enterprise plan (MCP is an Enterprise feature).
• An org admin to enable MCP + expose projects (one-time, below).
• Your own WrenAI account (for the OAuth login). No API key, no proxy.

---

Part 1 · Admin: enable MCP for your organization

(One-time, done by an organization admin.)

1. In WrenAI, go to Settings → Organization → MCP.
2. Turn on Enable MCP for this organization.
3. Under Exposed projects, toggle Exposed on for each project you want reachable over MCP. Optionally set New projects default to Exposed.
4. Copy the endpoint URL shown here — it's the same for everyone in the org: https://cloud.getwren.ai/api/mcp

Only exposed projects are reachable, and each user still only sees projects they’re a member of. Toggling a project off here (or disabling the org switch) immediately cuts MCP access to it.

---

Part 2 · Connect your AI client

You only need the endpoint URL and your WrenAI login. Pick your client:

claude.ai (Web) & Claude Desktop

1. Settings → Connectors → Add custom connector.
2. Name: WrenAI · URL: https://cloud.getwren.ai/api/mcp
3. Connect → you’ll be redirected to log in with your WrenAI account (OAuth). Approve access.

ChatGPT

(Requires connector/Developer-Mode access on your ChatGPT plan; exact menu names vary by version.)

1. Open Plugins → Connectors and start a new connector (New Plugin).
2. Name: WrenAI. Keep Connection: Server URL and enter https://cloud.getwren.ai/api/mcp.
3. Authentication: OAuth → tick “I understand and want to continue” → Create.
4. Log in with your WrenAI account when prompted.

Claude Code (CLI)

claude mcp add --transport http wrenai https://cloud.getwren.ai/api/mcp
# then, inside Claude Code:
/mcp        # select "wrenai" and complete the OAuth login in your browser

No proxy and no API key — you connect to the remote endpoint and authenticate via OAuth.

If /mcp hangs at the OAuth step, update Claude Code to the latest version and retry (older versions don't fully support remote OAuth MCP servers). Please open an issue here if it persists.

Self-hosted WrenAI

Use your own instance’s origin instead of cloud.getwren.ai, e.g. https://wren.yourcompany.com/api/mcp. Everything else is identical.

---

Part 3 · Using it

Once connected, just ask in natural language. Behind the scenes the agent:

1. calls list_projects → picks the right projectId,
2. reads schema with get_project_metadata,
3. answers with ask (or generate_sql / run_sql), and renders a generate_chart visualization.

Example prompts:

• “List my WrenAI projects.”
• “In the Sales project, what were the top 5 products by revenue last quarter? Chart it.”

Multiple orgs? If your account belongs to more than one MCP-enabled org, list_projects shows the exposed projects across all of them.

Optional: the WrenAI Analyst skill

For sharper analysis behavior, install the wrenai-analyst skill:

Client	How
claude.ai / Desktop	Customize → Skills → Create → Upload web/claude/wrenai-analyst/SKILL.md
ChatGPT	Paste web/chatgpt/wrenai-analyst-skill.md into Custom Instructions
Claude Code	claude plugin marketplace add ./ && claude plugin install wrenai-mcp

Generate the skill files: git clone https://github.com/canner/wrenai-mcp.git && cd wrenai-mcp && ./setup --host web (or --host chatgpt).

---

Tools

Tool	Purpose	Takes projectId
list_projects	Projects you can query	—
get_project_metadata	Table schema for a project	✅
ask	End-to-end natural-language Q&A	✅
generate_sql	Question → SQL	✅
run_sql	Execute SQL	✅
generate_chart	Vega chart spec	✅
generate_summary	Summarize results	✅

Data security & limits

• Per-user RLS/CLS — row/column policies are resolved server-side from the Wren AI account that authenticated the connection (the OAuth user); the client can’t override them. ⚠️ Avoid shared accounts — anyone using a shared Wren AI account inherits that account’s data permissions, so RLS/CLS can’t tell the individuals apart. Have each person connect with their own account.
• Enterprise-only.
• Interactive mode only — Agentic Mode is not available over MCP.

---

Migrating from the per-project URL

The legacy per-project URL — https://cloud.getwren.ai/api/mcp/{token} — and its Settings → Project → MCP connection page are deprecated and will be removed soon. To migrate:

1. Have an admin enable MCP + expose the project in Settings → Organization → MCP (Part 1).
2. In your AI client, replace the old per-project URL with https://cloud.getwren.ai/api/mcp and reconnect (Part 2).
3. Your agent now picks the project via list_projects instead of the URL.