Skip to main content
Glama
Misar-AI

MisarReach MCP Server

README.md
# MisarReach MCP Server

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

[![npm](https://img.shields.io/npm/v/@misarreach/mcp)](https://www.npmjs.com/package/@misarreach/mcp)
[![smithery](https://img.shields.io/badge/smithery-misar%2Fmisarreach--mcp-blue)](https://smithery.ai/server/misar/misarreach-mcp)
[![license](https://img.shields.io/badge/license-MIT-green)](./LICENSE)

**31 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)

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

### Claude Code

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

### Manual (any client)

```json
{
  "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/`](./connectors).

### Remote (no install)

```json
{
  "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:

1. **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`.
2. **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 leads already saved to the account, newest first, with paging and search. |
| `search_leads` | Start an AI lead-search job and return its jobId immediately. |
| `get_search_job_status` | Poll one lead-search job for its progress and results. |
| `submit_lead_feedback` | Record whether an AI-generated outreach message for a lead was good or bad, as training signal for future generations. |
| `discover_companies` | Find COMPANIES matching firmographic criteria, optionally pulling contact emails for each. |
| `enrich_lead` | Fill in a saved lead's missing person and company detail — seniority, department, LinkedIn, phone, company size, industry. |
| `verify_emails` | Check whether email addresses are deliverable, one or up to 20 at a time. |
| `score_leads` | Queue AI qualification scoring for leads — either every unscored lead in a search job, or a specific set of ids. |
| `list_lead_lists` | List the managed lead lists this account has created or synced. |
| `create_lead_list` | Create a new, empty managed lead list. |
| `sync_lead_list` | Import a managed lead list into local lead records so the rest of these tools can work with it. |
| `preview_message` | Draft a sample AI-personalised outreach message for a named person, to show what the agent would say. |
| `send_to_campaign` | Add saved leads to a campaign's contact list in bulk, up to 500 at a time. |
| `list_deals` | List deals as a flat, paged array, optionally filtered by status, with revenue totals alongside. |
| `create_deal` | Open a new deal against a lead's email address. |
| `update_deal` | Change a deal's status, value, or notes. |
| `get_pipeline` | Get the sales pipeline as a board: deals grouped by stage, with revenue totals. |
| `move_deal_stage` | Move one deal to a different pipeline stage — the equivalent of dragging its card on the board. |
| `start_autopilot` | Start an autonomous outreach run: give the agent a goal and it finds, contacts and follows up with leads on its own. |
| `list_autopilot_runs` | List past and running autopilot runs with their status and result summaries. |
| `get_autopilot_status` | Get the current progress and results of one autopilot run. |
| `get_channels_status` | Report the configuration, connection state and delivery stats for every outreach channel — WhatsApp, SMS and push. |
| `update_channel` | Turn one outreach channel on or off — WhatsApp, SMS or push. |
| `get_sales_agent_config` | Fetch the AI sales agent's current settings: whether it is enabled, its booking link, offer price, reply limits and confidence threshold. |
| `update_sales_agent_config` | Change the AI sales agent's settings — enable or disable it, set the booking link, offer price, reply limits and confidence threshold. |
| `get_sales_agent_actions` | Get what the AI sales agent has done today, with summary stats — actions taken, deals created, replies sent. |
| `process_sales_agent` | Run the sales-agent pipeline over one conversation: decide the next action and carry it out. |
| `search_verified_leads` | Find people or businesses matching a plain-language description and return only leads with a real way to reach them (email, phone or LinkedIn), each read from a real record — never generated. |
| `reveal_company_contacts` | Read a company's own website now and return what it publishes: contact emails, a phone number, its social profiles and its detected tech stack. |
| `company_email_format` | The email address format a company uses (e.g. |
| `verify_email` | Check whether one email address can receive mail. |

## 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/`](./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.

---

## Links

- Website — https://www.misarreach.com
- App — https://reach.misar.io
- Documentation — https://docs.misar.io/reach/mcp
- Smithery — https://smithery.ai/server/misar/misarreach-mcp
- npm — https://www.npmjs.com/package/@misarreach/mcp
- Source — https://github.com/Misar-AI/misarreach-mcp

MIT © [Misar.AI](https://misar.io)

TDQS

A3.9/5.0

Scored across 34 tools

Disambiguation3/5

Several tools overlap in purpose: search_leads, search_verified_leads and discover_companies all find leads/companies, while verify_email and verify_emails differ only by batch size. Descriptions cross-reference each other extensively, which helps, but the boundaries still require careful reading.

Naming Consistency4/5

Nearly all names use snake_case with a verb_noun pattern (list_leads, create_deal, update_channel, get_autopilot_status). Minor deviations like login, logout, auth_status and the noun-style company_email_format are conventional and do not impair readability.

Tool Count2/5

With 34 tools, the surface is well above the 25-tool threshold for 'too many'; although the platform spans several sub-domains, some tools (e.g. verify_email/verify_emails, search_leads/search_verified_leads) could be consolidated. The count feels heavy and increases selection cost.

Completeness3/5

Core lead, deal and agent workflows are covered, but there are notable gaps: campaigns are referenced by send_to_campaign yet no tool creates, lists, or manages campaigns, and there is no way to stop an autopilot run once started. Delete operations for leads, deals, and lists are also absent.

Maintenance

ActivityMaintained
ResponsivenessNo issues