MCP 服务器

MisarReach MCP

MCP server for MisarReach that enables finding, enriching, and scoring leads, running multi-channel outreach (WhatsApp, SMS, push) with AI autopilot, and managing the sales pipeline across 27 tools, 5 prompts, 4 resources, and 4 agent skills.

README

MisarReach MCP Server

Find leads, enrich and score them, run multi-channel outreach, and manage the sales pipeline — from any AI assistant.

27 tools · 5 prompts · 4 resources · 4 agent skills

Works with Claude (Desktop, Code, and web), Cursor, VS Code, Windsurf, Cline, Zed, Gemini CLI, ChatGPT, and any other MCP-compatible client — over stdio or Streamable HTTP.

Install

Smithery (recommended)

npx -y @smithery/cli install misar/misarreach-mcp --client claude

Claude Code

claude mcp add misarreach -- npx -y @misarreach/mcp@latest

Manual (any client)

{
  "mcpServers": {
    "misarreach": {
      "command": "npx",
      "args": ["-y", "@misarreach/mcp@latest"],
      "env": { "MISARREACH_API_KEY": "mrk_your_key_here" }
    }
  }
}

Ready-made configs for every client live in connectors/.

Remote (no install)

{
  "mcpServers": {
    "misarreach": {
      "type": "streamable-http",
      "url": "https://api.misar.io/reach/mcp",
      "headers": { "Authorization": "Bearer mrk_your_key_here" }
    }
  }
}

Authentication

Two options — no copy-paste needed for the first:

Browser login. Start the server with no key and run the login tool. It opens the MisarReach consent screen, you review the requested permissions, and the key is delivered straight back and saved to ~/.misarreach/config.json.
API key. Create one at https://reach.misar.io/settings/api-keys and set MISARREACH_API_KEY.

Self-hosted instances: set MISARREACH_BASE_URL.

Tools

Tool	Description
`list_leads`	List saved leads for the authenticated user.
`search_leads`	Start an AI-powered lead search job.
`get_search_job_status`	Poll the status and results of a lead search job..
`submit_lead_feedback`	Submit positive or negative feedback on an AI-generated lead message to improve future output quality..
`discover_companies`	Discover companies matching criteria via Hunter.io.
`enrich_lead`	Enrich a lead with person and company data (seniority, department, LinkedIn, phone, company size, industry).
`verify_emails`	Verify email deliverability for one or more email addresses.
`score_leads`	Trigger on-demand AI scoring for all unscored leads in a job, or a specific set of lead IDs.
`list_lead_lists`	List Hunter.io lead lists associated with the account..
`create_lead_list`	Create a new Hunter.io lead list..
`sync_lead_list`	Sync a Hunter.io lead list into local lead records..
`preview_message`	Generate a sample AI-personalised outreach message for a person.
`send_to_campaign`	Bulk-import selected lead IDs into a campaign's contact list.
`list_deals`	List deals with optional status filter.
`create_deal`	Create a new deal linked to a lead email..
`update_deal`	Update a deal's status, value, or notes..
`get_pipeline`	Get the full sales pipeline board with deals grouped by stage (new, contacted, interested, meeting, proposal, closed, lost) plus revenue totals..
`move_deal_stage`	Move a deal to a new pipeline stage (drag-and-drop equivalent)..
`start_autopilot`	Start an AI autopilot run for outreach automation.
`list_autopilot_runs`	List past autopilot runs with their status and summary results..
`get_autopilot_status`	Get the current status and progress of a specific autopilot run..
`get_channels_status`	Get the status, configuration, and delivery stats for all outreach channels (WhatsApp, SMS, push notifications)..
`update_channel`	Enable or disable a specific outreach channel (whatsapp, sms, or push)..
`get_sales_agent_config`	Fetch the current user's AI sales agent configuration..
`update_sales_agent_config`	Update the AI sales agent configuration (enable/disable, cal link, offer price, reply limits, confidence threshold)..
`get_sales_agent_actions`	Fetch today's AI sales agent actions and summary stats (total actions, deals closed, flagged for human review)..
`process_sales_agent`	Run the AI sales agent pipeline on a conversation — decides the next action and optionally generates a reply.

Prompts

Reusable workflows your client exposes as slash-commands.

Prompt	Description
`build_lead_list`	Find, enrich, verify and score a target lead list before any outreach.
`outreach_sequence`	Design and preview a multi-channel sequence, sending only after approval.
`pipeline_review`	Review deal pipeline health and recommend the next actions.
`autopilot_audit`	Check what autopilot has been doing and whether it should keep running.
`sales_agent_tuning`	Review the AI sales agent's configuration and recent actions.

Resources

Read-only context an agent can attach without spending a tool call.

URI	Description
`misarreach://channels`	Which outreach channels are connected and healthy.
`misarreach://lead-lists`	Your saved lead lists with sizes, so a campaign targets a real audience..
`misarreach://pipeline`	Current pipeline by stage — the baseline for any performance question..
`misarreach://autopilot/status`	Whether autopilot is running, and its current sending posture..

Agent skills

Bundled in skills/ — guidance an agent loads when a task matches.

Skill	Use when
`audit-autopilot-and-agent`	Check what MisarReach autopilot and the AI sales agent have been doing, and whether they should keep running. Use when asked to review automation, autopilot, or agent behaviour.
`build-and-qualify-a-lead-list`	Find companies matching an ICP, enrich and verify the contacts, score them, and save a qualified list. Use when asked to find leads, build a prospect list, or research a target market in MisarReach.
`review-the-sales-pipeline`	Report deal pipeline health and recommend next actions. Use when asked about deals, pipeline, forecast, or sales performance in MisarReach.
`run-an-outreach-sequence`	Design, preview and send a multi-channel outreach sequence. Use when asked to run a campaign, start outreach, or message a lead list in MisarReach.

Safety

Destructive and irreversible actions are annotated (destructiveHint) so clients can prompt before running them. The skills instruct agents to confirm before anything that sends mail, publishes content, or is otherwise visible to other people.

Discovery (initialize, tools/list, prompts/list, resources/list) never requires credentials, so registries can index the server without one. Every action does.