MCP server

The D50 MCP server

Give any AI agent 100+ real actions across sales, marketing, content, and hiring. The D50 Model Context Protocol server exposes the entire platform — the same tools a human uses in the app — to Claude, Cursor, or any MCP-compatible client.

Streamable HTTPapp.d50.ai/api/mcp100+ toolsBearer + OAuth1 credit = $1

The server speaks the MCP 2025-03-26 Streamable-HTTP transport. One endpoint, JSON-RPC over HTTPS:

Method	Path	Purpose
`POST`	`/api/mcp`	JSON-RPC requests (initialize, tools/list, tools/call)
`GET`	`/api/mcp`	SSE stream for server notifications
`DELETE`	`/api/mcp`	Terminate the session
`GET`	`/api/mcp/health`	Health check (no auth)

Connect a client

Claude Desktop / Claude Code

Add D50 to your MCP config with an API key from app.d50.ai → Settings → API keys:

{
  "mcpServers": {
    "d50": {
      "type": "http",
      "url": "https://app.d50.ai/api/mcp",
      "headers": { "Authorization": "Bearer d50_live_…" }
    }
  }
}

Claude.ai (custom connector)

Add https://app.d50.ai/api/mcp as a custom connector and authorize via OAuth — D50 implements the MCP OAuth 2.0 flow (/api/mcp/oauth/*), so no manual key handling.

Cursor / any MCP client

Point the client at the same HTTP URL with the Authorization: Bearer header. The agent calls tools/list and discovers every tool automatically.

Authentication

Bearer API keys — every call carries Authorization: Bearer d50_…. Keys are created per-agent in the app, stored only as a SHA-256 hash, and can carry an expiry.
Tool scopes — a key can be unrestricted (*) or limited to a named set of tools, so an agent only gets the actions you grant it.
OAuth 2.0 — for hosted clients (Claude.ai), the standard authorization-code flow is supported.
Sessions — stateful, 30-minute TTL; team membership is validated at auth time so multi-tenant data stays correctly scoped.

Same wallet, same rate. Every tool call bills the unified credit wallet (1 credit = $1) — an AI agent pays exactly what a human pays. Paid plans include 300 (Pro) to 1,000 (Agency) requests/min.

Tool catalog — 100+ tools

The full surface of the platform, grouped by capability — a representative selection per category below. Every tool is callable over MCP and the REST API.

CRM · selected tools

create_crm_contactsearch_crm_contactsupdate_crm_contactcreate_dealupdate_deallist_dealsmark_deal_wonmark_deal_lostcreate_pipelinelist_pipelinesget_crm_statsadd_notelog_activity

Lead Generation · selected tools

search_leadsreveal_leadenrich_leadget_leadget_lead_filters

Communication & Outreach · selected tools

send_emailsend_smssend_whatsappsend_whatsapp_templatemake_callmake_web_callinitiate_ai_callget_call_statusget_communication_channels

Content & Media · selected tools

generate_content_scriptgenerate_video_scriptcreate_contentgenerate_captioncreate_videogenerate_imageedit_imagegenerate_logogenerate_static_adgenerate_ad_copyrender_videocreate_lipsyncupload_mediarecommend_video_model

Scraping & Research · selected tools

scrape_websitecrawl_websitescrape_linkedin_companyscrape_instagram_profilescrape_tiktok_profilescrape_youtube_channelscrape_job_boardextract_structured_datalocal_business_searchgoogle_search

Funnels & Pages · selected tools

create_funnelcreate_funnel_stepupdate_funnelpublish_funnelgenerate_ai_funnelclone_landing_pageclone_websitelist_funnel_analytics

SEO & Competitive Intel · selected tools

seo_keyword_volumeseo_related_keywordsseo_discover_new_keywordsseo_keyword_difficultyseo_run_backlinks_sweepcompetitor_seo_intelcompetitor_keyword_overlapseo_run_llm_mentions_sweepanalyze_ad_creativescan_competitor_ads

Knowledge & Memory · selected tools

save_knowledgesearch_knowledgeget_task_briefingsemantic_searchlog_episoderecall_similar_episodesadd_watch_targetscan_watch_target

AI Agents · selected tools

create_agentdeploy_agentlist_agentsget_agent_statusstop_agentget_agent_activity_log

Recruitment · selected tools

create_candidatelist_candidatesinitiate_screening_callschedule_ai_interviewupdate_candidate_stagereject_candidate

Calendar & Booking · selected tools

book_calendar_meetingcheck_calendar_availabilitylist_calendar_meetingslist_meetings

Ads · selected tools

list_meta_campaignsget_meta_campaign_insightspause_meta_campaignresume_meta_campaignupdate_meta_campaign_budget

YouTube Pipeline · selected tools

yt_create_pipeline_configyt_trigger_pipeline_runyt_watchlist_addyt_discover_channelsyt_get_run_status

Billing & Account · selected tools

get_credit_balanceget_credit_historyget_account_statusget_analyticsqb_create_invoiceqb_list_invoices

Example agent session

An agent finds prospects, reveals contacts, and starts an AI call — three tool calls, one wallet:

// 1) find prospects
search_leads({ industry: "saas", country: "AE", limit: 25 })
// 2) reveal the best one's contact (bills 0.30 credits)
reveal_lead({ leadId: "lead_…" })
// 3) let an AI agent call and book a meeting (bills per minute)
initiate_ai_call({ to: "+9715…", goal: "book_meeting", agent: "sdr" })

That is the difference between an AI that can talk and an AI that can act. Start free and point your agent at D50.