Available to all organizations — Tenzo hosts a Model Context Protocol server that exposes the read endpoints of the Tenzo Public API as tools that any MCP-compatible AI client can call, plus a
submit_feedback write tool the assistant uses to log questions it struggled with (internal feedback only).Overview
Tenzo’s MCP server lets you point an MCP-compatible client (Claude, ChatGPT, Microsoft Copilot, Gemini Spark, Cursor, Windsurf, etc.) at your Tenzo workspace and ask natural-language questions like “summarize the last 20 calls for job X” or “list applications stuck in screening for more than a week.” The model can call Tenzo tools directly to fetch the data it needs. There are two ways to connect:- OAuth connector (recommended) — sign in with your Tenzo account; no API key to manage. Works with hosted clients like Claude, ChatGPT, Microsoft Copilot, and Gemini Spark that support the MCP HTTP transport with OAuth 2.1, and with local clients (Claude Code, Cursor, Windsurf) that support remote OAuth servers. Use this whenever you can — access follows your Tenzo login and permissions automatically.
.mcp.jsonwith an API key — a fallback for clients that can’t do the OAuth flow, or for users without a Tenzo account (e.g. a service or automation authenticating with an API key instead of a personal login).
https://api.tenzo.ai/mcp. Jump to your client: Claude · ChatGPT · Microsoft Copilot · Gemini Spark · other local clients.
The server is read-only for your recruiting data: the tools that proxy the Public API only expose GET endpoints. Write operations on recruiting data (creating jobs, upserting applications, posting comments) are not available through MCP — use the Public API directly for those. The one exception is submit_feedback, which lets the assistant record questions it struggled with so we can improve it; it writes only to an internal feedback log and exposes no recruiting data.
Option 1 — OAuth connector (Claude, ChatGPT, Microsoft Copilot, Gemini Spark)
Tenzo is published in the Claude, ChatGPT, and Microsoft Copilot connector directories, so you add it by finding Tenzo in the client’s connector list and clicking connect — no URL to paste, no API key. For Gemini Spark, add Tenzo as a custom app by pasting the MCP URL (Gemini does not currently offer a public directory submission for third-party connectors). The first time you use it, you sign in to Tenzo and approve access. The steps differ slightly per client.Claude (Claude.ai and Desktop)
A connector added to your Claude account works across Claude.ai, Claude Desktop, and mobile.1
Open the connector directory
Go to Customize → Connectors, click the +, and choose Browse connectors (or open claude.ai/directory).
2
Add Tenzo
Find and select Tenzo, then click Connect.
3
Approve access
Claude opens Tenzo’s sign-in page. Log in with your normal Tenzo account, then click Approve on the consent screen.Tenzo now appears in Claude’s tools list and is ready to use.
ChatGPT
Tenzo is a published ChatGPT app, so you just connect it from the directory — no Developer mode required (that’s only for custom MCP server URLs). Available on Plus, Pro, Business, Enterprise, and Edu plans.1
Open the apps directory
In ChatGPT, open Plugins from the sidebar (or Settings → Plugins, formerly “Apps”).
2
Connect Tenzo
Find Tenzo in the directory and click Connect.
3
Approve access
ChatGPT opens Tenzo’s sign-in page. Log in with your Tenzo account and click Approve on the consent screen. In a chat, open + → More to use the Tenzo tools; approve individual actions when the approval card appears.
On Enterprise and Edu workspaces (and any workspace where apps are restricted), a workspace admin must enable Tenzo under Workspace settings → Apps before members see the Connect button.
Microsoft Copilot
Tenzo is a published Power Platform connector, so you add it to a Copilot Studio agent from the connector list — no server URL to enter.1
Add a connector tool
Select your agent, open its Tools page, and choose Add a tool → Connector.
2
Select Tenzo
Search for Tenzo, select it, and pick the tool(s) you want to add. The Add tool pane opens.
3
Create a connection
Select Create new connection and sign in with your Tenzo account to approve access, then select Submit and Add and configure. Connections use each end user’s credentials by default.
4
Publish the agent
Tenzo’s read-only tools now appear on the agent’s Tools tab. Publish the agent and set its availability to make the tools reachable for end users in Microsoft 365 Copilot.
Tenzo is a premium connector, so it requires a Copilot Studio / Power Platform plan that includes premium connectors. A newly certified connector takes ~10–14 days to roll out to all regions after it goes live.
Gemini Spark (custom app)
Gemini Spark can call Tenzo through Custom apps for Spark — paste Tenzo’s MCP URL; there is no official Gemini Connected Apps directory listing for Tenzo today. Google currently requires a personal Google Account with Gemini Spark / Google AI Ultra eligibility (availability varies by region and account type), Keep Activity on, and a desktop browser at gemini.google.com to add the custom app. After you connect once on web, Spark can use the same connection on mobile.1
Open Connected Apps
Go to gemini.google.com → Settings & help → Connected Apps (under Personal Intelligence if needed).
2
Add Tenzo as a custom app
Under Custom apps for Spark, choose Add a custom app and enter:Leave Advanced credentials empty — Tenzo supports Dynamic Client Registration and Google’s Client ID Metadata Document flow, so no client ID or secret is required.
3
Approve access
Gemini opens Tenzo’s sign-in page. Log in with your Tenzo account, then click Allow on the consent screen (the screen shows the Gemini / Google client name and hostname).
4
Use Tenzo in Spark
In a Spark chat, select Tenzo (or mention it with
@) when you want Spark to query your recruiting data. Access is scoped to your Tenzo organization and permissions.To disconnect, unlink or remove the custom app under Gemini Connected Apps (or manage linked apps in your Google Account). Tenzo issues short-lived access tokens only — no refresh tokens — so after unlink, Gemini cannot mint a new token without consent again.
Revoking access
For Claude, ChatGPT, and Gemini Spark, Tenzo issues short-lived access tokens only — no refresh tokens. Once the access token expires, the client has to send the user back through the OAuth flow to get a new one. To stop a connector from talking to Tenzo, remove it from the client (Customize → Connectors in Claude, Settings → Plugins in ChatGPT, Connected Apps in Gemini); any in-flight token will stop working when it expires, and the client cannot mint a new one without going through consent again. Microsoft Copilot uses a confidential client that can refresh tokens, so revoke access by deleting the connection in Copilot Studio (or removing the connector from the agent).Option 2 — .mcp.json with an API key (fallback)
Prefer the OAuth flow above whenever your client supports it — it needs no key and scopes access to your Tenzo login. Use this API-key method only when OAuth isn’t an option: a client that can’t do the remote OAuth flow, or a user or service without a Tenzo account authenticating with an API key instead.
Local clients read MCP server definitions from a .mcp.json (or equivalent) config file. Add the following entry:
1
Generate a Tenzo API key
Follow Generating an API Key to create a key in Settings → Developer Tools. Your key starts with
tenzo_pk_.2
Export the key
Export it as an environment variable so it stays out of your config file:
3
Add the entry to your client config
- Claude Code — add to
.mcp.jsonat the root of your project (or your global Claude Code config) - Cursor —
~/.cursor/mcp.json - Windsurf —
~/.codeium/windsurf/mcp_config.json - Claude Desktop —
claude_desktop_config.json(see Anthropic’s docs for the platform-specific path)
4
Restart the client
Restart your MCP client. The Tenzo tools should appear in the client’s tool picker.
Available tools
Every tool maps 1:1 to aGET endpoint in the Tenzo Public API. Click any tag link below to see the full request and response schemas for the underlying endpoint.
Analytics
Applications
ATS
Audit
Calls & interviews
Candidates
Jobs
Folders
Media
Meeting Copilot
Docs
Organization & users
Templates
Two tools cover every org-template family. Pass the family via thekind parameter instead of calling a per-family tool. Both map to the Templates tag in the Public API.
Allowed
kind values: script_generation, background_info, redaction, suggested_follow_up, voicemail_sms_prompt, flagged_event, summary, thank_you_email, custom_booking_outro, rejection_email, resume_feedback_email, interview_results_layout, question_constraints, work_accommodation_review.
The REST API keeps a dedicated endpoint per template family (e.g.
GET /api/v1/templates/summary) — only the MCP tool surface is consolidated into list_templates / get_template.Troubleshooting
A tool returns 403 Forbidden
A tool returns 403 Forbidden
The tool was called with an ID that belongs to a different organization. MCP tools only return data scoped to your org. Double-check the ID is one your account can see in the Tenzo dashboard.
A tool I expected isn't in the list
A tool I expected isn't in the list
The MCP tools that proxy the Public API cover only its
GET endpoints. Write operations on recruiting data (e.g. POST /api/v1/job) must be called via the REST API directly. (The assistant also has a submit_feedback tool that writes only to an internal feedback log, not to your recruiting data.)Related
- API & Webhook Configuration — generate the API key used by
.mcp.jsonclients - Tenzo Public API Documentation — full schemas for every underlying endpoint
- Model Context Protocol specification — protocol details