Outladder MCP server

https://outladder.com/mcp

Streamable HTTP. Sign-in is OAuth 2.1 with dynamic client registration, so most apps only need the address. Your own code can use a workspace API key instead.

1. In Claude (web or desktop), open Settings, then Connectors.
2. Choose Add custom connector. Name it Outladder and paste https://outladder.com/mcp.
3. Press Connect. Outladder opens. Sign in, pick the workspace and what Claude may do, then press Allow.
4. In a chat, turn the Outladder connector on and ask, for example, "who replied this week?".

Claude Code:

claude mcp add --transport http outladder https://outladder.com/mcp

Then run /mcp in Claude Code and choose Authenticate.

1. In ChatGPT, open Settings, then Apps and Connectors. Custom connectors may need developer mode on under Advanced settings, depending on your plan.
2. Create a connector. Name it Outladder, paste https://outladder.com/mcp and choose OAuth for authentication.
3. Sign in to Outladder when asked, pick the workspace and what ChatGPT may do, then press Allow.

Menu names change from time to time. If a step looks different, look for "connector" or "MCP" in the settings.

An admin makes a key on Settings, API keys with the scopes mcp:read, and if needed mcp:write and mcp:sensitive. Send it as a bearer token:

curl https://outladder.com/mcp \
  -H "Authorization: Bearer gtm_..." \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

Each key has its own rate limit (set on the key, 120 calls a minute by default). A sign-in through an app gets 120 a minute. Every tool call is in the audit log.

Read. Always on. These run under your workspace's data rules and never change anything:

• search_accounts: Search accounts
• get_account: Open an account
• search_contacts: Search contacts
• get_contact: Open a contact
• list_signals: Signals
• list_sequences: Sequences
• list_replies: Replies
• list_meetings: Meetings
• meetings_by_signal: Meetings by signal
• list_deals: Deals
• list_content: Posts
• get_analytics: Funnel
• reply_rates: Reply rates
• search_context: Context library
• list_pending_actions: Approval inbox
• workspace_status: Workspace status
• list_lists: Lists

Ask for changes (write). Each call becomes a pending action. On autopilot it runs within your limits. On approve each it waits in the inbox. Enrollment still checks every contact and every email still goes through policy:

• create_list: Make a list
• enroll_contacts: Enroll in a sequence
• pause_sequence: Pause a sequence
• resume_sequence: Start a sequence again
• draft_email: Draft an email
• draft_social_post: Draft a social post
• draft_blog_post: Start a blog post

Sensitive changes. Need their own permission when you connect, and always wait for a person to approve in Outladder's inbox, whatever your autonomy settings:

• launch_workspace: Launch Outladder
• pause_all_agents: Pause all agents
• resume_all_agents: Resume all agents
• connect_tool: Connect a tool
• disconnect_tool: Disconnect a tool

Connecting a tool never passes keys or passwords through the AI app. Outladder sends you a link to finish it yourself.

• An app only sees the workspace you picked, as you, with your role.
• Write tools take an optional idempotency_key, so a retry never asks twice.
• Text in your records (emails, signals, company pages) comes from outside. Outladder labels it as data. Your AI app should treat it the same way.
• Disconnect an app any time on Settings, AI apps. Revoke a key on the API keys page.