Connect ReachRobin to your AI tool
ReachRobin exposes an MCP (Model Context Protocol) endpoint so any compatible AI tool - Claude Code, the Claude apps, ChatGPT, Cursor, Windsurf, or your own agent - can read your data and act on your behalf. Connect by signing in (recommended — no token to manage) or with a token. Both are below.
https://mcp.reachrobin.com/api/mcpSign in (recommended — no token)
Most MCP clients now let you sign in directly. Add the endpoint above and your tool opens a browser asking you to sign in to ReachRobin and approve access — there’s nothing to copy or paste, and no token to rotate. Access is tied to your ReachRobin account and the credential refreshes automatically.
claude.ai & the Claude desktop app
Open Settings → Connectors → Add custom connector, paste https://mcp.reachrobin.com/api/mcp, and click Connect. A ReachRobin sign-in window opens — approve it and the tools appear. No token, no JSON.
Claude Code (CLI)
Add the server with no auth header, then authenticate in the browser:
claude mcp add --transport http reachrobin https://mcp.reachrobin.com/api/mcp
Then run /mcp inside Claude Code and choose Authenticate— it opens your browser to sign in. (If you added it with an Authorizationheader, it stays on that token instead and won’t prompt.)
ChatGPT & other MCP clients
Add a custom connector pointing at https://mcp.reachrobin.com/api/mcp with no auth header. Any client that supports MCP sign-in (OAuth) discovers the flow automatically and prompts you to log in. If your client only accepts a static header, use a token instead (below).
Or connect with a token
Prefer a static credential — for automation, or a client that only supports an Authorization header? Create a token and paste it in.
1. Create a token
Sign in to ReachRobin and go to Settings → MCP tokens. Click New token, give it a label, and hit Create. The raw token is shown exactly once. Copy it before closing the dialog.
2. Paste the token into your AI tool
Right after creation, we show two ready-to-paste JSON snippets — pick the one that matches your tool. Tokens are prefixed rr_mcp_.
Claude Code (one-liner CLI — easiest)
If you use the Claude Code CLI, skip the JSON. Run one of these — pick the scope you want:
Available in all your projects (written to your home config).
claude mcp add --scope user --transport http reachrobin \ https://mcp.reachrobin.com/api/mcp \ --header "Authorization: Bearer YOUR_TOKEN"
ChatGPT (custom connector)
In ChatGPT, open Settings → Connectors → Add custom connector. Paste https://mcp.reachrobin.com/api/mcp as the URL and Bearer YOUR_TOKEN as the authorization header. No JSON editing required.
Cursor / Windsurf / Claude Code .mcp.json
Native Streamable-HTTP clients. Drop this into ~/.cursor/mcp.json, project-level .cursor/mcp.json, or .mcp.json for Claude Code. The type: "http" field is required by Claude Code (Cursor accepts it harmlessly):
{
"mcpServers": {
"reachrobin": {
"type": "http",
"url": "https://mcp.reachrobin.com/api/mcp",
"headers": { "Authorization": "Bearer YOUR_TOKEN" }
}
}
}Claude Desktop
Claude Desktop’s claude_desktop_config.json is stdio-only — it does not accept a bare url field for remote servers. To reach ReachRobin from the desktop app you bridge through the mcp-remote npm package, which runs locally as a stdio MCP server and proxies to our HTTPS endpoint. You need Node.js installed:
{
"mcpServers": {
"reachrobin": {
"command": "npx",
"args": [
"-y",
"mcp-remote",
"https://mcp.reachrobin.com/api/mcp",
"--header",
"Authorization:Bearer YOUR_TOKEN"
]
}
}
}Note the colon-without-space in Authorization:Bearer — spaces inside argsaren’t always escaped by Claude Desktop.
Alternatively, if you use claude.ai in the browser, add ReachRobin as a Custom Connector under Settings → Connectors — point it at https://mcp.reachrobin.com/api/mcp and paste the token when prompted; no JSON editing required.
Restart the AI tool after saving. You should see the ReachRobin tools appear in its tool list. So far we’ve verified the Cursor snippet against Cursor v1.0+ and the Claude Desktop snippet via mcp-remote; other clients (Windsurf, Zed, custom agents) generally accept one of these two shapes.
Available tools
However you connect, you have access to 60tools, grouped by domain. Every tool respects your plan limits and role — if your AI tries something you don’t have permission for, you’ll get a clear error back.
Identity
whoamiIdentity echo
Outreach campaigns
add_clients_to_campaignAdd clients (companies) to a campaignarchive_outreach_campaignArchive (delete) an outreach campaigncreate_outreach_campaignCreate an outreach campaignget_outreach_campaignGet an outreach campaignget_outreach_campaign_clientsList enrolled companies in a campaignget_outreach_campaign_pending_recommendationsList AI drafts awaiting approval in a campaignget_outreach_campaign_progressGet campaign enrichment + sequence progressget_outreach_campaign_timelineGet campaign event timelinelist_outreach_campaignsList outreach campaignspause_outreach_campaignPause an outreach campaignremove_clients_from_campaignRemove clients from a campaignresume_outreach_campaignResume an outreach campaignset_campaign_sequenceSwap the sequence template of a campaign (requires feature.sequencing.ui)trigger_campaign_enrichmentForce-refresh enrichment for a campaignupdate_outreach_campaignUpdate name / brand of an outreach campaign
Prospect campaigns
get_prospect_campaignGet a prospect campaignlist_prospect_campaignsList prospect-research campaignslist_prospect_leadsList prospect leads in a campaignlist_prospect_to_outreach_candidatesList prospect leads ready to convert
Clients
get_clientGet a clientget_client_activityGet a client activity feedlist_clientsList clients/prospectslist_recommendations_historyList the user message recommendations history
LinkedIn accounts
get_linkedin_account_healthGet LinkedIn account health and recent eventslist_linkedin_accountsList connected LinkedIn accounts
Brands & sequences
get_brandGet a brand profileget_brand_sequenceGet the default sequence for a brand (requires feature.sequencing.ui)get_sequenceGet a sequence detail (requires feature.sequencing.ui)list_brandsList brand profileslist_sequencesList sequences (requires feature.sequencing.ui)
Conversations & messaging
approve_pending_recommendationApprove and send an AI-drafted DMget_conversationGet a conversation threadlist_conversationsList inbox conversationssend_dmSend a DM in an existing LinkedIn conversation
Plan & usage
get_plan_infoGet current plan + entitlementsget_rate_limit_statusGet current rate-limit usagelist_top_upsList top-up history
Other
canvas_archiveMCP: archive a campaign canvascanvas_attach_channelMCP: attach a LinkedIn channel to a canvascanvas_attach_sourceMCP: attach an audience source to a canvascanvas_commitMCP: commit a new canvas definitioncanvas_create_from_briefMCP: create a campaign canvas from a briefcanvas_getMCP: read a campaign canvascanvas_get_historyMCP: get campaign canvas event historycanvas_get_statsMCP: get campaign canvas statscanvas_launchMCP: launch a campaign canvascanvas_listMCP: list campaign canvasescanvas_pauseMCP: pause a campaign canvascanvas_resumeMCP: resume a paused campaign canvasconnect_integrationConnect an integration (LinkedIn/Google/Outlook, HubSpot, WhatsApp, email)disconnect_integrationDisconnect an existing integration connectionedit_recommendationEdit the drafted message of a pending recommendationgenerate_recommendationsGenerate fresh recommendations for a clientlist_integrationsList all integration connections and their statuslist_pending_recommendationsList the actionable pending recommendations queuereconnect_integrationRe-authenticate an existing integration connectionreject_recommendationReject a pending recommendation (optionally regenerate)undo_recommendationUndo an approved recommendation (cancels the pending send)verify_email_domainRe-check verification status of a bring-your-own sending domain
Rate Limiting
- 200 requests per minute per token (shared across all tool calls).
- Invitations are not sent directly via MCP. They flow through campaign automation.
- Applicable limitson messaging, enrichment, and invitations depend on your plan and any top-ups — not the MCP layer. The quotas that apply to you are the same ones that apply to your account in the app.
When you hit a limit the tool returns 429 with a resetAt timestamp. Your AI can wait and retry automatically.
Privacy & security
- Sign-in (OAuth) connectionsdon’t hand your tool a long-lived secret — it holds a short-lived access token, refreshed automatically, scoped to your ReachRobin account. Remove the connector in your AI tool to disconnect it. The notes below apply to token connections.
- We store only a hash of the token — the raw value is shown once at creation and is unrecoverable afterwards. Lose it? Revoke and create a new one.
- Tokens expire after 90 days.You’ll see the expiry date on the tokens list; rotate before then.
- Revoke any token instantly from the tokens page in Settings. Revocation takes effect on the next authenticated request — typically within a few minutes.
- Every MCP call is logged with the token id, tool name, status, and duration. If you suspect a compromised token, revoke it immediately and contact support.