Skip to content

Claude integration

Connect Claude to Tallyfy

Connect Claude to Tallyfy and you can manage your work just by chatting. Ask Claude about your tasks, launch a process, or clean up a template, and it does the work in your Tallyfy account for you. It works in both Claude.ai on the web and the Claude Desktop app, on any paid Claude plan.

You connect once by adding Tallyfy as a connector in Claude’s settings, then signing in with your Tallyfy account. Claude takes it from there. The exact steps for the web app and the desktop app are under For developers.

What you can do

Once it’s connected, you talk to Claude in plain language and it works in Tallyfy for you. Claude can:

  • Handle your tasks - find your open tasks, create new ones from a sentence, update them, and mark them done.
  • Run and track processes - launch a workflow from a template, check on running work, and archive what’s finished.
  • Build and tidy templates - read a template, add steps, suggest deadlines and form fields, and check a template’s health for problems.
  • Manage form fields and automation rules - add or change fields, and create or clean up the if-then rules that drive your workflows.
  • Work with comments and people - read and add task comments, and look up or invite team members and guests.

How to connect

The setup is quick. In Claude’s settings, open Connectors (or Integrations in the desktop app), add a custom connector named Tallyfy pointing at the Tallyfy MCP server1, and sign in with your Tallyfy account to authorize it. Add it once and it syncs across Claude on the web, desktop, and mobile. Step-by-step clicks for both apps are under For developers.

What it costs

You need a paid Claude plan (Pro, Max, Team, or Enterprise). There’s no extra Tallyfy charge for using this, but the tool calls count toward your Anthropic usage limits.

Is it secure?

Yes. Anthropic doesn’t train on your data. The Claude Desktop app stores your sign-in token in your computer’s secure keychain (macOS Keychain or Windows Credential Manager), never in plain text. Each organization’s data stays separate, every interaction is logged for audit, and Anthropic maintains SOC 2 Type II compliance. The sign-in detail is under For developers.

For developers

(Skip this unless you’re setting up the technical side.)

The integration is live. The Tallyfy MCP server at https://mcp.tallyfy.com supports OAuth 2.1 with PKCE. Use Streamable HTTP transport - SSE is no longer supported. Both Claude.ai (web) and Claude Desktop connect through it, and the connection uses OAuth 2.1 with PKCE2 - Claude handles the auth flow automatically when you add the server URL.

Set up Claude.ai (web)

Available on paid Claude plans (Pro, Max, Team, Enterprise). Once added, the connector syncs across Claude web, Claude Desktop, and Claude Mobile.

  1. Open Claude.ai settings

    Go to claude.ai and click your profile icon in the top-left corner, then select Settings.

  2. Go to Connectors

    In Settings, click Connectors in the left sidebar.

  3. Add a custom connector

    Click Add custom connector. Enter:

    • Name: Tallyfy
    • URL: https://mcp.tallyfy.com

    Then click Save.

  4. Authorize with your Tallyfy account

    Claude.ai opens a browser window to Tallyfy’s authorization page. Sign in with your Tallyfy credentials and click Authorize to grant Claude access to your organization.

  5. Verify the connection

    Start a new conversation. The Tallyfy connector appears in your active integrations. Try:

    What are my open tasks in Tallyfy?

    Claude calls get_my_tasks and returns your current task list.

Set up Claude Desktop

Available on paid Claude plans (Pro, Max, Team, Enterprise). Download for macOS or Windows.

  1. Install or update Claude Desktop

    Download from claude.ai/download. Remote MCP server support with OAuth requires a recent version.

  2. Open settings

    Launch Claude Desktop, then open Settings. On macOS use Cmd+, or click the Claude menu - then Settings. On Windows use the gear icon or the application menu.

  3. Go to Integrations

    Go to the Integrations tab (labelled Extensions or MCP servers in some versions).

  4. Add the Tallyfy MCP server

    Click Add integration, Add remote server, or the + button. Enter:

    https://mcp.tallyfy.com

    Claude Desktop discovers the OAuth endpoints automatically - no manual configuration needed.

  5. Authorize with your Tallyfy account

    Claude Desktop opens a browser window to Tallyfy’s authorization page. Sign in and click Authorize. The token is stored in your OS keychain (macOS Keychain or Windows Credential Manager) - never in plaintext.

  6. Restart Claude Desktop

    Quit completely (Cmd+Q on macOS) and relaunch. This ensures the MCP connection initializes cleanly.

  7. Verify the connection

    Open a new conversation. Look for the tools icon at the bottom of the chat input - clicking it shows available Tallyfy tools. Try:

    What are my open tasks in Tallyfy?

    Claude calls get_my_tasks and returns your current task list.

How the integration works

Tallyfy’s MCP server runs on Cloudflare at https://mcp.tallyfy.com and exposes your Tallyfy data as tools Claude can call during a conversation.

The flow works like this:

  1. Tool discovery - Claude connects and queries the MCP server for available tools
  2. OAuth 2.1 authentication - Claude handles the auth flow. You sign in with Tallyfy credentials and authorize access
  3. Tool calls - When you ask a question, Claude picks the right Tallyfy tool and passes the correct parameters
  4. API execution - The MCP server uses your credentials to call the Tallyfy API and returns structured results
  5. Response - Claude formats the data into a natural language answer

Authentication architecture

The MCP server uses OAuth 2.1 with PKCE (S256). All discovery and token exchange happens through https://mcp.tallyfy.com, which proxies to Tallyfy’s authorization server:

EndpointURL
MCP serverhttps://mcp.tallyfy.com
Protected resource metadatahttps://mcp.tallyfy.com/.well-known/oauth-protected-resource
Authorization server metadatahttps://mcp.tallyfy.com/.well-known/oauth-authorization-server
Authorization (proxy)https://mcp.tallyfy.com/mcp/oauth/authorize
Token exchange (proxy)https://mcp.tallyfy.com/mcp/oauth/token
Client registration (proxy)https://mcp.tallyfy.com/mcp/oauth/register
JWKS (token verification)https://mcp.tallyfy.com/.well-known/jwks.json

Supported OAuth scopes:

ScopeAccess
mcp.tasks.readRead tasks and assignments
mcp.tasks.writeCreate, update, and complete tasks
mcp.processes.readRead workflow runs
mcp.processes.writeLaunch, update, and archive processes
mcp.templates.readRead templates and steps
mcp.templates.writeEdit templates, steps, and automations
mcp.forms.readRead form fields
mcp.forms.writeAdd and modify form fields
mcp.users.readRead organization members and guests
mcp.users.writeInvite and manage users
mcp.automation.readRead automation rules
mcp.automation.writeCreate and modify automations

MCP server tools

The Tallyfy MCP server exposes these tools to Claude:

Task tools

  • get_my_tasks - Get all tasks assigned to the current user
  • get_user_tasks - Get tasks for a specific team member by user ID
  • get_tasks_for_process - Get all tasks in a specific process by ID or name
  • create_task_from_text - Create a task from natural language - extracts title, deadline, and assigns to named members or guests
  • complete_task - Mark a task as complete
  • reopen_task - Reopen a previously completed task
  • update_task - Update task properties like title, deadline, or description

Search tools

  • search_for_tasks - Search tasks across the organization by keyword
  • search_for_processes - Search workflow processes/runs by keyword
  • search_for_templates - Search templates by keyword (use get_all_templates for a full list)

Process tools

  • get_organization_runs - Get all active workflow processes with optional filters for status, template, owner, tags, and type
  • launch_process - Launch a new process from a template with a name and optional tags/folders
  • get_process - Get details for a specific process by ID
  • update_process - Update process properties like name or tags
  • archive_process - Archive a completed or cancelled process

Template tools

  • get_template - Get a template by ID or name with full details
  • get_all_templates - Get all templates in the organization
  • get_template_steps - Get all steps for a template
  • get_kickoff_fields - Get kick-off form fields for a template
  • get_step_dependencies - Analyze which automations control when a step appears
  • assess_template_health - Retrieve complete template data for a full health assessment by the AI
  • suggest_step_deadline - Get deadline recommendations for a step based on its type and complexity
  • suggest_kickoff_fields - Get suggested kick-off fields based on template content
  • add_step_to_template - Add a new step to an existing template
  • add_assignees_to_step - Add member or guest assignees to a step
  • edit_description_on_step - Update the description/instructions on a step
  • update_template - Update template properties like title or description
  • clone_template - Create a copy of an existing template with a new name

Form field tools

  • add_form_field_to_step - Add text, dropdown, date, or other field types to a step
  • update_form_field - Update an existing form field’s properties
  • move_form_field - Reorder a form field within a step
  • delete_form_field - Remove a form field from a step
  • get_dropdown_options - Get options for a dropdown field
  • update_dropdown_options - Update dropdown choices
  • suggest_form_fields_for_step - Get AI recommendations for useful form fields based on step content

Automation tools

  • create_automation_rule - Create if-then automation rules on a template
  • update_automation_rule - Modify an existing automation rule’s conditions or actions
  • delete_automation_rule - Remove an automation rule
  • analyze_template_automations - Analyze all automations with redundancy detection (duplicates, same-trigger groups)
  • get_step_visibility_conditions - See which automations control a step’s visibility with actual condition details
  • suggest_automation_consolidation - Get prioritized recommendations for consolidation (orphaned rules, duplicates, conflicts, merge candidates)

Comment tools

  • get_task_comments - Get all comments on a specific task
  • add_task_comment - Add a comment to a task

User and guest tools

  • get_organization_users - Get all organization members with profile data
  • get_organization_users_list - Get members with minimal data for listing
  • get_organization_guests - Get guest users with full data
  • get_organization_guests_list - Get guests with minimal data
  • invite_user_to_organization - Invite a new member by email

Organization tools

  • get_groups - Get groups in the organization
  • get_folders - Get folders for organizing templates and processes
  • get_tags - Get tags used across the organization

Example prompts

Template health audit

Prompt: "Assess the health of our Customer Onboarding template and
suggest improvements to the automation rules."
Claude calls: assess_template_health -> analyze_template_automations ->
suggest_automation_consolidation, then summarizes findings and
recommended fixes.

Process launch and tracking

Prompt: "Launch the Employee Onboarding template for Jane Doe and
show me all the tasks that were created."
Claude calls: launch_process -> get_tasks_for_process, then lists
every task with assignees and deadlines.

Writing effective prompts for Claude

Be specific about what you want analyzed and how you want the output structured:

Good: "Analyze the proposal considering technical feasibility, financial
viability, strategic alignment, and risk factors. Provide weighted scoring
for each dimension with detailed justification."
Bad: "Review the proposal and approve or reject."

Give Claude explicit reasoning steps:

Good: "First, identify all people affected. Second, analyze impact
on each group. Third, recommend mitigation strategies. Finally, provide
an implementation timeline."
Bad: "Figure out who this affects and what to do."

Security and compliance

  • No training on your data - Anthropic doesn’t train on API inputs
  • JWT validation - Tokens validated via RS256 using Tallyfy’s public key. Only Tallyfy can issue valid tokens
  • OS keychain storage - Claude Desktop stores tokens in macOS Keychain or Windows Credential Manager, never in plaintext
  • Data isolation - Each organization’s data stays separate
  • Audit logging - Complete interaction history
  • SOC 2 Type II - Anthropic maintains compliance

Troubleshooting

“MCP tools icon doesn’t appear”

  • Restart Claude Desktop completely (Cmd+Q on macOS, not just closing the window)
  • Check Settings - then Integrations to confirm https://mcp.tallyfy.com is listed and active
  • If the server shows an error, remove and re-add it

“Authorization fails or loops”

  • Confirm you’re signing in with the Tallyfy account that has access to the target organization
  • Open https://mcp.tallyfy.com/.well-known/oauth-protected-resource in your browser - a JSON response means the server is reachable
  • Clear Claude Desktop’s stored credentials and re-authorize from scratch

“SSE connection errors”

  • SSE transport isn’t supported anymore. The Tallyfy MCP server uses Streamable HTTP only
  • Update Claude Desktop to the latest version, which defaults to Streamable HTTP

Best practices

  1. Start with read-only tools - Use get_my_tasks, search_for_processes, and get_all_templates first to get familiar with the data before running write operations
  2. Be specific - Name the template, process, or user explicitly. Claude performs best with complete context
  3. Add human checkpoints - For write operations like creating automation rules, review Claude’s plan before confirming
  4. Iterate on prompts - Review Claude’s outputs and refine your instructions over time

Footnotes

  1. Model Context Protocol - Anthropic’s open standard letting AI models call external tools and access data securely

  2. Proof Key for Code Exchange - prevents authorization code interception attacks during the OAuth flow