Skip to content
Tallyfy

Tallyfy MCP server with Google Gemini

Tallyfy connects to Google’s three Gemini surfaces through OAuth 2.1 with streamable-HTTP transport. For developers, the Gemini CLI and Gemini Code Assist auto-discover OAuth via Dynamic Client Registration against https://mcp.tallyfy.com/. For enterprise teams, admins add Tallyfy as a Custom MCP Server data store in the Gemini Enterprise console; we recommend using https://mcp-gcp.tallyfy.com/ (the Cloud Run mirror) for that path since it puts the origin inside Google’s own network.

Gemini MCP support status

  • Gemini CLI - remote streamable-HTTP MCP via ~/.gemini/settings.json; OAuth auto-discovered through DCR and PKCE
  • Gemini Code Assist - same configuration schema as Gemini CLI in VS Code; IntelliJ uses mcp.json
  • Gemini Enterprise - Custom MCP Server data store (Preview); per-organization configuration in the Google Cloud Console
  • Apigee integration - optional path for converting managed APIs into MCP servers (not needed for Tallyfy; we publish our own server directly)

Prerequisites

  • A Tallyfy account with an active organization
  • For the Gemini CLI path: @google/gemini-cli installed (npm install -g @google/gemini-cli)
  • For Gemini Code Assist: the Code Assist extension installed in VS Code or IntelliJ, signed in to a Google account with Gemini access
  • For Gemini Enterprise: Discovery Engine Editor IAM role (roles/discoveryengine.editor) plus the ability to override the org policy that blocks custom MCP server data stores

Connect Tallyfy in Gemini CLI

  1. Install the CLI

    Terminal window
    npm install -g @google/gemini-cli
    gemini --version

  2. Add Tallyfy to your settings

    Edit ~/.gemini/settings.json (or .gemini/settings.json inside a project):

    {
      "mcpServers": {
        "tallyfy": {
          "httpUrl": "https://mcp.tallyfy.com/"
        }
      }
    }

    That’s the entire config. The CLI handles OAuth automatically through Dynamic Client Registration plus PKCE the first time Tallyfy responds with a 401. You don’t need to paste a client ID, client secret, or any OAuth endpoint URLs.

  3. Restart Gemini CLI

    Quit any running session, then re-launch:

    Terminal window
    gemini

  4. Verify the connection

    Inside the Gemini chat, run:

    /mcp list

    You’ll see tallyfy listed with 115 tools.

  5. First use authorizes you

    The first prompt that calls a Tallyfy tool opens a browser tab where you sign in to Tallyfy and pick which organization to connect.

Connect Tallyfy in Gemini Code Assist

  1. VS Code

    Add the same mcpServers block to ~/.gemini/settings.json. Reload the Gemini Code Assist extension (Cmd+Shift+P then Developer: Reload Window). Then run /mcp list in the chat panel.

  2. IntelliJ

    Create mcp.json in the IDE’s Gemini config directory with the same content. Restart the IDE and verify with /mcp list.

  3. Same OAuth flow

    Both IDEs use the same Dynamic Client Registration flow. Your browser opens for sign-in the first time you call a Tallyfy tool.

Connect Tallyfy in Gemini Enterprise

Gemini Enterprise’s Custom MCP Server data store (Preview as of May 2026) is configured per-organization by your Google Cloud admin. The connection setup happens entirely on Google’s side. There’s no approval step or vendor signup at Tallyfy.

  1. Override the org policy

    Custom MCP server data stores are blocked by default at the organization level. Your admin overrides the policy following Google’s setup documentation ↗.

  2. Grant Discovery Engine Editor

    Assign roles/discoveryengine.editor to the admin who’ll create the data store, via the IAM page in Cloud Console.

  3. Register an OAuth client for Tallyfy

    Tallyfy supports Dynamic Client Registration. Register a confidential client for Gemini Enterprise with one curl call. The server advertises both client_secret_basic and client_secret_post as supported token endpoint auth methods - Gemini Enterprise typically uses client_secret_basic, but either works:

    Terminal window
    curl -s -X POST https://mcp-gcp.tallyfy.com/mcp/oauth/register \
      -H "Content-Type: application/json" \
      -d '{
        "client_name": "Gemini Enterprise - YourOrgName",
        "redirect_uris": ["https://vertexaisearch.cloud.google.com/oauth-redirect"],
        "grant_types": ["authorization_code","refresh_token"],
        "response_types": ["code"],
        "token_endpoint_auth_method": "client_secret_basic"
      }'

    The response contains both client_id and client_secret. Save both - you’ll paste them into the Cloud Console in the next step. You can register against either endpoint (mcp.tallyfy.com or mcp-gcp.tallyfy.com); the credentials are valid on both because they share the OAuth backend.

  4. Add the data store in Cloud Console

    In Google Cloud Console: Gemini Enterprise -> Data stores -> Create data store -> Custom MCP Server (Preview) -> Add MCP server. Fill the form using the Cloud Run mirror so your origin stays inside Google’s network:

    FieldValue
    MCP Server URLhttps://mcp-gcp.tallyfy.com/
    Authorization URLhttps://mcp-gcp.tallyfy.com/mcp/oauth/authorize
    Token URLhttps://mcp-gcp.tallyfy.com/mcp/oauth/token
    Client ID(from step 3 response)
    Client Secret(from step 3 response)
    Token endpoint auth methodclient_secret_basic (or client_secret_post)
    Scopesmcp.users.read mcp.tasks.read mcp.tasks.write mcp.processes.read mcp.processes.write mcp.templates.read mcp.templates.write mcp.forms.read mcp.forms.write mcp.automation.read mcp.automation.write

    If you prefer the DigitalOcean origin, swap all three URLs to mcp.tallyfy.com - the behavior is identical.

  5. Add a server description

    In the MCP Server Description field, paste this starter (Google uses it for routing decisions):

    Tallyfy manages business workflows, processes, and task assignments. Use this server when the user asks about:
    - Running, launching, or tracking workflow processes
    - Assigning, completing, or commenting on tasks
    - Creating or editing process templates and form fields
    - Managing approval flows and automation rules
    - Viewing process performance or finding bottlenecks
    

    Example queries that route here:
    - "Show me overdue tasks in Tallyfy"
    - "Launch a new vendor onboarding process for ACME Corp"
    - "What's the average completion time for our customer support workflow?"
    - "Add a new step to the invoice review template"

    Adjust this for your team’s vocabulary.

  6. Finalize and reload

    Select a Multi-region location, give the data connector a name, and click Create. Wait for the state to change to Active. Then go to Actions -> Reload custom actions and enable the individual Tallyfy tools your team should use (tools are disabled by default).

Practical examples

Task management via natural language

Prompt:

Using Tallyfy, find all overdue tasks and create a summary grouped by assignee.

Gemini calls search_for_tasks with an overdue filter, groups by assignee, and writes a summary.

Process analytics

Prompt:

Analyze our customer onboarding processes from the last month and identify bottlenecks.

Gemini queries process history through MCP tools, computes per-step completion times, and surfaces the slowest steps.

Template review

Prompt:

Review our Invoice Processing template and suggest improvements.

Gemini fetches the template with get_template, walks the step dependencies and automation rules, and proposes changes.

Gemini-specific features

Google Cloud integration

Because Gemini runs inside Google’s platform, you can combine Tallyfy data with other Google services:

  • BigQuery - analyze Tallyfy process data alongside your data warehouse
  • Cloud Run - deploy your own MCP logic at scale
  • Pub/Sub - drive event-based workflow triggers

Multimodal input

Gemini can process images and documents alongside text. That’s handy for visual inspection workflows or document review processes.

Vertex AI agents

Build multi-agent systems with the Agent Development Kit that combine Tallyfy with other services under Google’s governance controls.

Limitations

  • Preview status - Custom MCP Server data store is in Preview as of May 2026; the Console UI and field names may change.
  • Tools default to disabled - after creating the data store, you must manually enable each Tallyfy tool via Actions -> Reload custom actions.
  • No public Google directory for MCP server vendors - Tallyfy isn’t surfaced by default to Gemini Enterprise customers. Each org adds it themselves through the steps above.
  • Visual process tracker doesn’t render in chat - like other chat surfaces, Gemini shows you text. Complex forms become text interactions, and you’ll ask for status updates explicitly.
  • OAuth client types supported - Tallyfy supports both PKCE public clients (CLI, Code Assist) and confidential clients (Gemini Enterprise). Confidential clients can use either client_secret_basic or client_secret_post for the token endpoint; both methods are advertised in the server’s /.well-known/oauth-authorization-server metadata. All flows use OAuth 2.1 with S256 PKCE.

Best use cases

Data-driven optimization

Combine BigQuery analytics with Tallyfy process data:

Correlate customer support tickets from BigQuery with support process completion times in Tallyfy.

Multi-service automation

Coordinate across Google Cloud and Tallyfy:

When a new document lands in Cloud Storage, kick off the document review process in Tallyfy and notify the team via Gmail.

Process monitoring

Use Gemini’s analytical strengths for workflow insights:

Monitor active processes and alert me to any at risk of missing their SLA, with a root-cause hypothesis.

Security

  • Authentication - OAuth 2.1. Gemini CLI and Code Assist register as PKCE public clients through DCR. Gemini Enterprise registers as a confidential client through DCR using client_secret_basic or client_secret_post (both advertised by the server). All flows use S256 PKCE for code challenge. Refresh tokens rotate automatically.
  • Data handling - Google processes your data per their AI policies. Enterprise teams get Vertex AI controls and audit logging in Cloud Logging.
  • Network - All traffic over HTTPS. Enterprise deployments can add VPC Service Controls and private connectivity.
Get in touch
About Tallyfy