> ## Documentation Index > Fetch the complete documentation index at: https://parabola.io/docs/llms.txt > Use this file to discover all available pages before exploring further. # Connecting Parabola to AI tools with MCP > Use Parabola's MCP server to query your workspace, flows, and run metadata from Claude, ChatGPT, Glean, and other MCP-compatible AI tools. The Parabola MCP is available to customers on our Team, Advanced, and Enterprise plans. If your organization is on one of these plans and you'd like access, reach out to your account manager or [help@parabola.io](mailto:help@parabola.io). ## What you can do with the Parabola MCP 🔒 **Scoped to your permissions.** The MCP only sees flows you have editor or viewer access to in Parabola — if you can't see it in the app, the MCP can't either. Org admins and content admins have the same full access through the MCP that they already have in Parabola. Here are example questions the MCP handles well, organized by the use case they support. ### 1. Adoption, cost, and ROI visibility Answers the questions leadership keeps asking — about who's actually using Parabola and what the team has built. * *Are we on track on credit usage this billing period?* * *How many of our users are actually building?* * *Our three-way match flow saves 1 hour per run at \$25/hour — calculate the total dollar savings since it started running, and break it down by month if you can.* ### 2. Operational health monitoring Surfaces failing, paused, or at-risk flows across your workspace without clicking through each one. * *Give me a health check on our Parabola workspace — anything I should be worried about?* * *Which flows have failed at least once in the last 30 days, and what's the failure rate on each?* * *What's our overall flow creation rate month-over-month? Are we building more or less than we did six months ago?* ### 3. Coverage planning and single-point-of-failure detection Identifies what's at risk when someone is out, and where manual processes create fragility. * *Sarah is out next week. Pull up the flows she owns and flag any that don't have an automated trigger, but has run in the last 14 days.* * *Which flows in our team do not have an automated schedule?* * *Joanna is leaving the company at the end of the month. Show me her flows that have an active trigger and have run in the last 30 days — these are her production automations we need to find a new owner for before she leaves.* ### 4. Discoverability and reuse before building something new Finds existing flows and patterns before duplicating work. * *We're building a new reconciliation flow for a new entity — do we already have reconciliation logic I should look at?* * *Show me all flows owned by the finance/billing people (Paul, Andrew, Emory, me) with "billing" or "invoice" in the name.* * *Show me Ana's most recently created flows — they're probably working on something adjacent to what I'm building.* *** ## Server details | | | | -------------- | -------------------------------------------------------------------------------------------------------------- | | Server URL | `https://parabola.io/api/mcp` | | Authentication | OAuth 2.0. Actions run on behalf of the authenticated user and are scoped to that user's Parabola permissions. | | Transport | Streamable HTTP | | Rate limits | Applied per tool, per user | ## Before you connect These prerequisites apply regardless of which AI tool you're connecting from: 1. A Parabola account with access to the flows and data you want the AI to work with. 2. Permission (or admin approval) in your AI tool to add custom MCP connectors. Most enterprise plans gate this to admins. 3. Network reachability from the AI provider's servers to `https://parabola.io/api/mcp`. Claude, ChatGPT, and Glean all broker the connection from their own cloud — not from your machine — so review your networking and firewall settings if you have strict egress policies. *** ## Connect from Claude Applies to: [Claude.ai](http://claude.ai/) (web), Claude Desktop, Claude mobile, Cowork ### One-time initial setup (Likely needs to be done by an owner of the Claude account.) 1. Click **Customize → Connectors** and click **+** 2. Scroll down to **Custom connectors** and click **+** to add a new custom connector 3. Name this custom connector "Parabola" 4. Enter `https://parabola.io/api/mcp` and save 5. After this, individual members can enable Parabola from their own settings. ### Enabling the Parabola MCP 1. Click **Customize → Connectors** 2. Search "Parabola" 3. Click to connect and complete the Parabola OAuth sign-in. That's it! Now you can use the Parabola MCP in Claude. 📄 [Official Claude docs](https://support.claude.com/en/articles/11175166-get-started-with-custom-connectors-using-remote-mcp) *** ## Connect from Claude Code Applies to: Claude Code (CLI) Unlike Claude.ai and Claude Desktop, Claude Code connects from your own machine rather than brokering through Anthropic's cloud. There's no one-time admin setup — each person adds the server and authenticates individually from their terminal. OAuth is always per-user; one teammate connecting does not connect it for anyone else. **Network note:** Because Claude Code reaches `https://parabola.io/api/mcp` directly from your machine, the cloud-egress consideration in [Before you connect](#before-you-connect) doesn't apply — but your local network does need to reach that URL. ### Adding and authenticating 1. Add the Parabola server to Claude Code: ```bash theme={null} claude mcp add --transport http parabola https://parabola.io/api/mcp ``` By default this registers the server for the current project directory only. Add `--scope user` to make it available across all your projects: ```bash theme={null} claude mcp add --transport http --scope user parabola https://parabola.io/api/mcp ``` 2. Start a Claude Code session: ```bash theme={null} claude ``` 3. Authenticate by running `/mcp`, selecting **parabola**, and completing the Parabola OAuth sign-in in the browser window that opens. Then return to the terminal. 4. Confirm the connection with `claude mcp list` (in the terminal) or `/mcp` (in a session). You should see **parabola** listed. That's it! 🎉 ### For teammates Each person needs to run both steps above — adding the server **and** authenticating — on their own machine, since server definitions and OAuth tokens are stored locally per user. If you want teammates to skip the `add` step, run the command with `--scope project` instead. This writes the server definition to a `.mcp.json` file in the project directory that you can commit to your repo. Teammates who pull the repo inherit the server automatically and only need to run `/mcp` to authenticate with their own account. ### A note on connectors vs. CLI If you've already connected Parabola as a [custom connector in Claude.ai](#connect-from-claude), it may appear in your `/mcp` list inside Claude Code with a `claude.ai` prefix. That connector is managed in the Claude app's settings, not the CLI — `claude mcp remove` won't affect it. The steps above add a separate, CLI-managed connection. *** ## Connect from Gemini CLI Applies to: Gemini CLI (`gemini`), Google's terminal AI agent. As of June 18, 2026, Gemini CLI is only available to organizations with a **Gemini Code Assist Standard or Enterprise license**, or paid Gemini / Gemini Enterprise Agent Platform API keys. If you're on a personal Google account (free, Google AI Pro, or Ultra), Gemini CLI stopped serving requests around that time — use [Antigravity CLI](#connect-from-antigravity-cli) instead. Like Claude Code, Gemini CLI connects to Parabola directly from your machine rather than brokering through Google's cloud. There's no one-time admin setup on the Parabola side — each person adds the server and authenticates individually from their terminal. OAuth is always per-user; one teammate connecting does not connect it for anyone else. ### Adding and authenticating You'll do everything in this section from your computer's **Terminal** app. On Mac, open Spotlight (⌘+Space), type "Terminal", and press Enter. On Windows, open **Windows Terminal** from the Start menu. **Before you start:** you'll need to be signed in to a Google account that your organization has assigned a **Gemini Code Assist Standard or Enterprise license**. If your team has already set up a **Gemini API key** or **Vertex AI** access for you to use with Gemini CLI, those work too — but Code Assist is the path most Parabola customers will take. Gemini CLI needs Node.js version 20 or higher. Check by pasting this into Terminal and pressing Enter: ```bash theme={null} node --version ``` If you see `v20.11.0` or higher, skip to the next step. Otherwise, download the LTS version from [nodejs.org](https://nodejs.org), install it, then close and reopen Terminal. Paste and press Enter: ```bash theme={null} npm install -g @google/gemini-cli ``` Takes about 30 seconds. Paste and press Enter: ```bash theme={null} gemini ``` You'll see a **Get started** screen asking how to authenticate. Select **Sign in with Google** (arrow keys + Enter) and complete sign-in in the browser window that opens. Return to Terminal — you should land on the Gemini CLI welcome screen. Type `/quit` and press Enter; we'll come back in a later step. If you see *"Your current account is not eligible for Gemini Code Assist for individuals,"* your Google account doesn't have a Code Assist license. See the troubleshooting note at the bottom of this section. Paste and press Enter: ```bash theme={null} mkdir -p ~/.gemini && touch ~/.gemini/settings.json ``` This creates a hidden folder called `.gemini` in your home directory and an empty `settings.json` file inside it. Nothing visible happens — that's expected. Paste this whole block into Terminal and press Enter: ```bash theme={null} cat > ~/.gemini/settings.json << 'EOF' { "mcpServers": { "parabola": { "httpUrl": "https://parabola.io/api/mcp" } } } EOF ``` To confirm it saved correctly, paste this and press Enter: ```bash theme={null} cat ~/.gemini/settings.json ``` You should see the same JSON printed back to you. Paste and press Enter: ```bash theme={null} gemini ``` Inside Gemini CLI, type `/mcp` and press Enter. You'll see **parabola** listed. Select it with arrow keys, press Enter, and choose **Authenticate**. A browser window opens for the Parabola sign-in — complete it and come back to Terminal. Type `/mcp` again. Parabola should now show as 🟢 Ready with a list of tools underneath. 🎉 Try asking *"List my Parabola flows"*. The first time it uses a Parabola tool, you'll be asked to approve. ### For teammates Each person needs to run these steps — installing Gemini CLI, editing their own `~/.gemini/settings.json`, and authenticating — on their own machine, since MCP configs and OAuth tokens are stored locally per user. ### Troubleshooting * **Failed to sign in. Your current account is not eligible for Gemini Code Assist for individuals** — your Google account doesn't have a Gemini Code Assist Standard/Enterprise license. Two options: * If your organization uses Gemini API keys or Vertex AI, restart `gemini` and pick that auth method instead. * Otherwise, use [Antigravity CLI](#connect-from-antigravity-cli) — it's Google's replacement for personal-account users and takes about 5 minutes to set up. * **`command not found: gemini`** — the install didn't complete or your Terminal doesn't know about the new command yet. Close Terminal completely and reopen it. If that doesn't fix it, ask IT for help with the install. * **`EACCES` or permission denied during `npm install`** — your laptop is blocking the install. Ask IT to install `@google/gemini-cli` globally. * **"This app is blocked" during Google sign-in** — your Google Workspace admin hasn't approved Gemini CLI. Ask them to allow it in the admin console. * **No MCP servers configured** — Gemini CLI didn't find your config. Run `cat ~/.gemini/settings.json` and verify the file exists and matches the JSON above. * **Unauthorized \[Auth Needed]** on the parabola row — the server is reachable but you haven't completed OAuth yet. Type `/mcp`, select **parabola**, and choose **Authenticate**. * **Server doesn't show up at all** — verify you used `httpUrl`, not `url` or `serverUrl`. 📄 [Official Gemini CLI docs](https://geminicli.com/docs/tools/mcp-server/) *** ## Connect from Antigravity CLI Applies to: Antigravity CLI (`agy`), Google's replacement for Gemini CLI on personal Google accounts (free, Google AI Pro, and Ultra). If your organization has a Gemini Code Assist Standard or Enterprise license, use [Gemini CLI](#connect-from-gemini-cli) above instead. Antigravity CLI works the same way as Gemini CLI — it connects to Parabola directly from your machine and OAuth is per-user — but the config file lives in a different location and uses a different key name. ### Adding and authenticating You'll do everything in this section from your computer's **Terminal** app. On Mac, open Spotlight (⌘+Space), type "Terminal", and press Enter. On Windows, open **Windows Terminal** from the Start menu. **Before you start:** you'll need a Google account. Any tier works — free, Google AI Pro, and Ultra are all supported. Paste this into Terminal and press Enter: ```bash theme={null} curl -fsSL https://antigravity.google/cli/install.sh | bash ``` Takes about a minute. If Terminal says `agy: command not found` after installation, close Terminal completely and reopen it, then continue. Paste and press Enter: ```bash theme={null} agy auth login ``` A browser window opens. Sign in with your Google account, then return to Terminal. Paste and press Enter: ```bash theme={null} mkdir -p ~/.gemini/config && touch ~/.gemini/config/mcp_config.json ``` This creates the folder `~/.gemini/config` and an empty `mcp_config.json` file inside it. Nothing visible happens — that's expected. Paste this whole block into Terminal and press Enter: ```bash theme={null} cat > ~/.gemini/config/mcp_config.json << 'EOF' { "mcpServers": { "parabola": { "serverUrl": "https://parabola.io/api/mcp" } } } EOF ``` Antigravity CLI uses `serverUrl` — not `url` or `httpUrl` like other MCP clients. Copy-pasting a config from Gemini CLI, Cursor, or VS Code won't work. To confirm it saved correctly, paste this and press Enter: ```bash theme={null} cat ~/.gemini/config/mcp_config.json ``` You should see the same JSON printed back to you. Paste and press Enter: ```bash theme={null} agy ``` Inside Antigravity CLI, type `/mcp` and press Enter. You'll see **parabola** listed (likely with an "Unauthorized \[Auth Needed]" note — that's expected). Select it with arrow keys, press Enter, and choose **Authenticate**. A browser window opens for the Parabola sign-in — complete it and come back to Terminal. Type `/mcp` again. Parabola should now show as connected with a list of tools underneath. 🎉 Try asking *"List my Parabola flows"* to see it in action. The first time it uses a Parabola tool, you'll be asked to approve. ### For teammates Each person needs to run these steps on their own machine, since MCP configs and OAuth tokens are stored locally per user. ### Troubleshooting * **`agy: command not found`** — the installer finished but your Terminal doesn't know about it yet. Close Terminal completely and reopen it, then try again. * **Installer fails with a permissions error** — your laptop is blocking the install. Ask IT to install Antigravity CLI on your machine. * **"This app is blocked" during Google sign-in** — your Google Workspace admin hasn't approved Antigravity CLI. Ask them to allow it in the admin console. * **No MCP servers configured** — the config file is empty or malformed. Run `cat ~/.gemini/config/mcp_config.json` and verify it matches the JSON above. * **Unauthorized \[Auth Needed]** on the parabola row — the server is reachable but you haven't completed OAuth yet. Type `/mcp`, select **parabola**, press Enter, and choose **Authenticate**. * **Server doesn't show up at all** — check that you used `serverUrl` (not `url` or `httpUrl`) and that the file lives at `~/.gemini/config/mcp_config.json` (not `~/.gemini/settings.json`). 📄 [Official Antigravity CLI docs](https://antigravity.google/docs/mcp) *** ## Connect from Cursor Cursor connects to `https://parabola.io/api/mcp` directly from your machine over Streamable HTTP, and authenticates each user individually via OAuth. There's no one-time admin setup — every person adds the server and signs in on their own machine. ### Adding the server 1. Open Cursor and navigate to Settings. 2. Go to **Tools & MCP** and click **New MCP Server**. This opens your `mcp.json` file. 3. Add the Parabola server: ```json theme={null} { "mcpServers": { "parabola": { "url": "https://parabola.io/api/mcp" } } } ``` 4. Save the file (`Cmd/Ctrl + S`). Cursor picks up the change immediately — there's no separate save button in the settings panel. ### Authenticating 1. Return to **Settings → Tools & Integrations**. The **parabola** entry appears with a status indicator and a **Connect** button, since the server requires OAuth. 2. Click **Connect** and complete the Parabola OAuth sign-in in the browser window that opens, then return to Cursor. 3. Confirm the entry shows a green status indicator and that Parabola's tools are listed. If they don't appear, check **Developer: Show Logs** in the Command Palette. Cursor enforces a limit of roughly 40 tools across all enabled MCP servers combined. If Parabola's tools don't all load, disable other MCP servers you're not actively using. ### For teammates Each person adds the server and authenticates on their own machine — server definitions and OAuth tokens are stored locally per user. To let teammates skip the manual add, commit an `mcp.json` with the Parabola entry to your project; teammates who pull it inherit the server and only need to complete the OAuth sign-in. *** ## Connect from ChatGPT Applies to: ChatGPT web (and mobile, once connected on web) Plans: Business, Enterprise, Edu. Custom MCP connectors with full tool support aren't available on Free or Plus. ### One-time initial setup (Likely needs to be done by an owner of the ChatGPT account.) 1. Go to **Workspace Settings** 2. Go to **Apps** and hit **Create** 3. Enable **developer mode** 4. Click **New app form** and hit **Create** * **Name:** Parabola * **Description:** Run and inspect Parabola flows from ChatGPT * **MCP Server URL:** `https://parabola.io/api/mcp` * **Authentication:** OAuth 5. The app should now appear in your "Drafts" 6. Publish this draft step so the others can use it. ### Enabling the Parabola MCP for your ChatGPT account 1. In ChatGPT, go to **Settings → Apps** 2. Search "Parabola" 3. Click **Connect** and complete the Parabola OAuth sign-in. ### Using the Parabola MCP in chats Start a new conversation, click **+**, then **More**, then enable the Parabola connector for this conversation. You'll need to re-enable it in each new chat. 📄 [Official OpenAI docs](https://developers.openai.com/apps-sdk/deploy/connect-chatgpt) *** ## Connect from Glean Please contact [help@parabola.io](mailto:help@parabola.io) if you'd like to connect the Parabola MCP to Glean. *** ## Connect from other MCP-compatible tools Parabola's MCP server follows the open MCP spec, so it works with any compliant client. You need three things: 1. The server URL: [https://parabola.io/api/mcp](https://parabola.io/api/mcp) 2. OAuth 2.0 3. Support for Streamable HTTP *** ## We want to hear from you The Parabola MCP keeps getting better thanks to feedback from people using it. Share thoughts, questions, and ideas at [help@parabola.io](mailto:help@parabola.io).