> For the complete documentation index, see [llms.txt](https://help.daasity.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://help.daasity.com/core-concepts/ai-analyst/connect-your-ai-mcp/connect-gemini-cli.md). # Connect Gemini CLI Connect Gemini CLI to Daasity via MCP to query governed semantic views from your terminal. Setup, verification, and troubleshooting steps. Connect Gemini CLI to your governed Daasity data using the Model Context Protocol (MCP). Once connected, you can discover semantic views and query your workspace metrics in natural language directly from your terminal, without writing SQL or leaving your shell. {% hint style="info" %} **Do this in your terminal, not the Connections screen.** Gemini CLI is configured through a local settings file on your machine. The Connections screen in Daasity is where you generate your token; everything after that happens in Terminal. {% endhint %} ### Prerequisites * **AI Analyst enabled** on your Daasity workspace. * **Gemini CLI** installed on your machine, or Node.js and npm available so you can install it. * **A Google AI Studio API key** configured in Gemini CLI, or a linked Google Cloud billing account. * Permission to edit files in your home directory (`~/.gemini/`). {% hint style="warning" %} **Keep your token private.** Your Daasity MCP token is specific to your workspace and grants read access to your data. Treat the token like a password. Anyone with the token can query your data until it expires or you revoke it. Never paste it into a shared document, a ticket, or a public repository. {% endhint %} ### Connect Gemini CLI to Daasity {% stepper %} {% step %} ### Step 1 — Generate your Daasity MCP token 1. In Daasity, open the **AI Analyst** tab. 2. Select **Connect Your AI**. 3. Open the **Manual token** tab. 4. Choose **Gemini CLI** from the tool list, then select **Generate Token**. 5. Copy both the **MCP server URL** and the **token**. You'll need both in Step 3. Your token looks like `dsy_mcp_pk_...`. Your MCP server URL is specific to your environment, so always use the one shown on your own Daasity screen rather than one copied from documentation or from a colleague. {% hint style="info" %} **Don't reuse a URL from documentation.** MCP URLs are environment-specific. Always copy yours from your own Daasity screen. {% endhint %} {% endstep %} {% step %} ### Step 2 — Install Gemini CLI If you don't already have Gemini CLI on your machine, install it: ```bash npm install -g @google/gemini-cli ``` Confirm the install: ```bash gemini --version ``` {% endstep %} {% step %} ### Step 3 — Add Daasity as an MCP server Gemini CLI reads its MCP configuration from `~/.gemini/settings.json`. Open that file in your editor and add a `daasity` entry under `mcpServers`, replacing both placeholders with the values you copied in Step 1: ```json { "mcpServers": { "daasity": { "url": "", "type": "http", "headers": { "Authorization": "Bearer ", "Accept": "application/json, text/event-stream" } } } } ``` If the file already contains other servers or settings, add the `daasity` block alongside them rather than replacing the file. If `~/.gemini/settings.json` does not exist yet, create it with the contents above. Each field is doing specific work, and the connection will fail without all three: {% hint style="warning" %} **Fresh installs only.** If you have never configured Gemini CLI before and have no existing settings to preserve, you can write the whole file in one command. This **overwrites** `~/.gemini/settings.json` and will remove any other MCP servers or preferences already configured there. {% endhint %} ```bash python3 -c 'import json, os; p=os.path.expanduser("~/.gemini/settings.json"); d={"security":{"auth":{"selectedType":"gemini-api-key"}},"mcpServers":{"daasity":{"url":"","type":"http","headers":{"Authorization":"Bearer ","Accept":"application/json, text/event-stream"}}}}; open(p,"w").write(json.dumps(d,indent=2))' ``` {% endstep %} {% step %} ### Step 4 — Verify the connection Before launching the CLI, confirm that Daasity accepts your credentials and headers. Run this directly in your terminal, substituting your own URL and token: ```bash curl -X POST \ -H "Authorization: Bearer " \ -H "Accept: application/json, text/event-stream" \ -H "Content-Type: application/json" \ -d '{"jsonrpc": "2.0", "method": "tools/list", "id": 1}' ``` > **Connection checkpoint:** You should receive a JSON payload listing available tools, including `find_views` and `query`. If you see a `Not Acceptable` error, your `Accept` header is malformed. Check it in both the curl command and `~/.gemini/settings.json`. > {% endstep %} {% step %} ### Step 5 — Launch Gemini CLI Start the CLI from your normal terminal shell: ```bash gemini ``` If you're on the Google AI Studio free tier, pin the session to a high-quota model so background agents don't exhaust your rate limit mid-session: ```bash gemini --model ``` Users with a paid Google Cloud billing account linked to AI Studio can run `gemini` on production Pro models without quota concerns. {% endstep %} {% step %} ### Step 6 — Run your first query At the interactive `>` prompt, ask Daasity to describe your catalog. Gemini CLI invokes the `find_views` tool, compiles the correct Snowflake SQL against your governed semantic views, and returns your workspace data. ``` > Use Daasity to list my available semantic views and metrics > What are my top 5 selling products according to Daasity? > Query Daasity for my order and order line revenue ``` {% hint style="success" %} **Connection checkpoint:** If Gemini returns semantic-view names from your own workspace, the MCP connection is working. Because answers come from governed semantic views, they apply your workspace's valid-order rules, revenue definitions, and reporting timezone, so they reconcile to your Daasity dashboards. {% endhint %} {% endstep %} {% endstepper %} ### Troubleshooting | Symptom | Solution | | -------------------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | `HTTP 406` or `JSON-RPC -32600` on connect | The `Accept` header is missing or altered. It must read exactly `application/json, text/event-stream` in `~/.gemini/settings.json`. | | `Not Acceptable` from the curl test in Step 4 | Same cause as above. Fix the header in the curl command first, then in your settings file. | | The CLI hangs, times out, then falls back to Google Web Search | The connection is attempting SSE transport. Confirm `"type": "http"` is set on the `daasity` server and that you are not passing `--transport sse`. | | Free-tier quota exhausted within minutes | Shell commands were typed at the `>` prompt. The CLI reads them as coding questions and spawns background subagents. Exit the prompt and run all flags from your primary terminal shell. | | `401` or `403` from the curl test | The token is expired or revoked. Return to **AI Analyst → Connect Your AI → Manual token** and generate a new one. | | Connection succeeds but no semantic views are returned | AI Analyst may not be fully provisioned on your workspace yet. Semantic views are generated when the connection is first established; allow time for that to complete, then retry. | ### Readiness checklist | Check | Confirmed when | | -------------------- | -------------------------------------------------------------------------------------------------------- | | Token generated | You copied both the MCP server URL and token from **Connect Your AI**. | | Server configured | `~/.gemini/settings.json` contains a `daasity` entry with `type`, `Authorization`, and `Accept` all set. | | Credentials accepted | The Step 4 curl test returns a tool list containing `find_views` and `query`. | | CLI connected | Gemini CLI launches without transport warnings. | | Data flowing | A prompt at `>` returns semantic-view names from your own workspace. | ### What's next * Connections are read-only, tenant-isolated, quota-capped, and revocable at any time from **Connect Your AI**. * A token expires after 30 days without use. If queries start failing with a `401`, generate a new token. * To connect a different client, see the other guides under **Connect Your AI / MCP**.