How do I use Prosyo MCP Server?

1. Click on "Deploy Server". 2. Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state. 3. In the chat, type @ followed by the MCP server name and your instructions, e.g., "@Prosyo MCP Server show my running campaigns, unread replies, and credits" That's it! The server will respond to your query, and you can continue using it as needed. Here is a step-by-step guide with screenshots.

Prosyo MCP Server

by ShubhamTuts

Overview Schema Source Related Servers Discussions Admin

Remote

1-Minute Quickstart

The hosted endpoint is plug-and-play. There is no API key to create, no developer token to sign up for, and no authentication header to add. The client config is the URL: https://app.prosyo.com/ai.

The first time Claude or Cursor connects, Prosyo opens an allow screen in the browser so the workspace can be chosen. That confirm is once. It stays out of the config file.

The same steps are on the live guide, with a card for each client.

Claude Desktop

Add this to claude_desktop_config.json. Full walkthrough: Connect Prosyo to Claude.

{
  "mcpServers": {
    "prosyo": {
      "url": "https://app.prosyo.com/ai"
    }
  }
}

Restart Claude Desktop. Turn Prosyo on in the chat and ask.

Cursor

Add this to .cursor/mcp.json in the project, or to the global Cursor MCP config.

{
  "mcpServers": {
    "prosyo": {
      "url": "https://app.prosyo.com/ai"
    }
  }
}

Reload MCP servers. Prosyo is available in Agent chat.

Windsurf

Add this to ~/.codeium/windsurf/mcp_config.json. Windsurf's remote field is serverUrl. Headers stay empty.

{
  "mcpServers": {
    "prosyo": {
      "serverUrl": "https://app.prosyo.com/ai"
    }
  }
}

VS Code and GitHub Copilot

Add this to .vscode/mcp.json. The top-level key is servers, and the type is http.

{
  "servers": {
    "prosyo": {
      "type": "http",
      "url": "https://app.prosyo.com/ai"
    }
  }
}

Smithery

Install into Claude from the verified Smithery listing:

npx @smithery/cli install hello-kmun/prosyo --client claude

Try it the moment it connects

"Give me the Prosyo overview: running campaigns, unread replies, and credits." → get_overview
"Draft a four-touch campaign for heads of talent at UK agencies. LinkedIn first, email on day five. Leave it as a draft." → create_campaign
"Show unread inbox threads and draft a reply. Do not send it." → list_inbox, draft_reply

Reads run immediately. Anything that sends, launches, enrolls, unlocks, researches, or spends credits stops and waits for a yes in chat, then confirm_action.

Related MCP server: Outbound Engine MCP

Architecture

The agent talks to one hosted endpoint. Prosyo runs the outbound engine behind it. Nothing is installed on the machine. Diagram on the live guide.

flowchart LR
  A["AI Client (Claude / Cursor)"] -->|HTTPS Streamable HTTP| B["https://app.prosyo.com/ai"]
  B --> C["Outbound Engine (LinkedIn + Cold Email + Unified Inbox)"]

Hop	What happens
AI Client (Claude / Cursor)	Claude, Cursor, Windsurf, or VS Code calls a tool from chat.
HTTPS (Streamable HTTP)	One remote connection. No stdio process. No API key header.
`https://app.prosyo.com/ai`	Prosyo's hosted MCP server. Manifest: `server.json`.
Outbound Engine	LinkedIn, cold email, and the unified inbox for this workspace.

Tools Directory

30 tools. Read tools return data immediately. Write tools that change who gets contacted, or that spend credits, prepare the action and finish only after confirm_action once the chat says yes. pause_campaign and save_intent_leads_to_list apply immediately.

Workspace · Campaigns · Prospecting · Inbox · Intent · Safety

The same directory, with the same anchors, is on the GitHub Pages guide.

Workspace

Tool	Type (Read/Write)	Description
`get_overview`	Read	Dashboard snapshot: running campaigns, replies, unread inbox, credits, and account health for this Prosyo workspace.
`get_entitlements`	Read	Current plan name, trial or paid status, LinkedIn and email seat counts, active campaign limit, and credit balance.

Campaigns

Tool	Type (Read/Write)	Description
`list_campaigns`	Read	List campaigns in this workspace. Optional status filter: draft, running, paused, completed, archived.
`get_campaign`	Read	Campaign detail: status, enrolled count, sequence step types, and connected seats.
`preview_campaign`	Read	Merge a real prospect into the campaign sequence so the messages can be read before launch.
`create_campaign`	Write	Create a draft campaign. Does not launch or enroll anyone. Modes: blank, preset (`preset_id`), ai (`ai_brief`), or intent (audience fills with people who show `intent_signals` and fit the ideal customer; each waits for review first). Stays a draft until a later launch is confirmed.
`launch_campaign`	Write	Prepare launching a draft campaign. Sending does not start until `confirm_action` after a yes in chat.
`pause_campaign`	Write	Pause a running campaign immediately so it stops queueing new sends. Idempotent.
`resume_campaign`	Write	Prepare resuming a paused campaign. Sending does not restart until `confirm_action` after a yes in chat.

Prospecting & Enrichment

Tool	Type (Read/Write)	Description
`search_prospects`	Read	Search people in this workspace by name, company, or email.
`list_lists`	Read	Prospect lists with people counts.
`get_prospect_research`	Read	A prospect's latest research (facts with sources) and personalised lines: icebreaker, connection note, LinkedIn message, email subject, and body.
`enroll_prospects`	Write	Prepare adding a list or specific people to a campaign. Does not enroll until `confirm_action` after a yes in chat.
`research_prospects`	Write	Prepare researching prospects: sourced facts plus a personalised icebreaker, connection note, LinkedIn message, and email. Costs credits per person; fresh research is reused free. Runs only after `confirm_action`.
`enrich_prospects`	Write	Prepare enriching prospects (`enrich_only`) or writing a personalised first line (`personalized_line`). Costs credits per person. Runs only after `confirm_action`.

Unified Inbox

Tool	Type (Read/Write)	Description
`list_inbox`	Read	Recent LinkedIn and email conversations. Optional search and unread filter.
`get_thread`	Read	One inbox conversation with recent messages.
`draft_reply`	Write	Draft an inbox reply. Returns text only and does not send. Use `send_inbox_reply` after the reply should go out.
`send_inbox_reply`	Write	Prepare sending a LinkedIn or email reply from Inbox. The message is not sent until `confirm_action` after a yes in chat.

Intent Intelligence

Tool	Type (Read/Write)	Description
`get_intent_summary`	Read	Intent overview: whether tracking is on, this month's included people used and left, today's deliveries, High-intent and Warm counts, and how many wait for review.
`list_intent_leads`	Read	People or companies showing buying signals, strongest first, each with intent score, ideal-customer fit, and why now. Locked people show only their signals until unlocked.
`get_intent_lead`	Read	One Intent person or company: every signal with its source and date, fit reasons, research status, and personalised lines.
`get_ideal_customer`	Read	The workspace's ideal customer rules (roles, industries, company size, locations, exclusions) and the High-intent and automatic delivery settings.
`list_playbooks`	Read	Playbooks and Intent-led campaign audiences: which signals they watch, Review first or Autopilot, matches, people added, and credits used this month.
`list_review_queue`	Read	People a playbook matched who are waiting for approval, or whose researched message is ready to review.
`get_intent_results`	Read	What Intent found and produced: signals by source, people delivered, researched, contacted, replied, meetings, customers, credits per reply, and the best signals and playbooks.
`save_intent_leads_to_list`	Write	Save unlocked Intent people to a prospect list (existing `list_id` or a new `list_name`). Their personalised lines go with them. Free. Locked people are skipped. Idempotent.
`unlock_intent_leads`	Write	Prepare unlocking locked Intent people. Uses the plan's included people first, then credits per person. Nothing is unlocked until `confirm_action` after a yes in chat.

Execution & Safety

Tool	Type (Read/Write)	Description
`approve_review`	Write	Prepare approving people waiting in the review queue (run ids from `list_review_queue`). Approval researches them for credits. The message is still visible before it is sent. Runs only after `confirm_action`.
`confirm_action`	Write	Run a launch, enroll, resume, inbox send, unlock, research, enrichment, or review approval after an explicit yes in this chat. Pass `confirmation_token` from the previous tool.

Official MCP Registry

Prosyo is listed in the Official MCP Registry as io.github.ShubhamTuts/prosyo version 1.0.1.

The registry is the community index of publicly accessible MCP servers. The Model Context Protocol organization on GitHub runs it collaboratively, backed by Anthropic, GitHub, Microsoft, and PulseMCP.


Official platform	Model Context Protocol Registry
Registry service	modelcontextprotocol/registry, the Go service behind the centralized index
Reference servers	modelcontextprotocol/servers, reference implementations maintained by the MCP steering group
Record format	Open `server.json` schema: reverse-DNS name, discovery metadata, and the remote execution path
Downstream use	Upstream metadata index that clients, integrations, and other registries can read
Live guide	shubhamtuts.github.io/prosyo-mcp-server

Prosyo's record points at the hosted Streamable HTTP endpoint https://app.prosyo.com/ai. The manifest declares no API key and no authentication headers.

Resource Links

Live guide: https://shubhamtuts.github.io/prosyo-mcp-server/
App & Platform: https://prosyo.com
Docs: https://prosyo.com/docs
Support: support@prosyo.com
Connect Claude: https://prosyo.com/docs/ai/connect-claude/
Smithery: https://smithery.ai/server/hello-kmun/prosyo
Official MCP Registry: io.github.ShubhamTuts/prosyo
Registry service: modelcontextprotocol/registry
Reference servers: modelcontextprotocol/servers
Source: https://github.com/ShubhamTuts/prosyo-mcp-server

This server cannot be deployed

One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

Maintenance

ActivityMaintained

ResponsivenessNo issues

Resources

Need Help?

Related Servers

Related MCP Connectors

campaignstack A
io.campaignstack
Run B2B outreach from your AI agent: 250+ tools for campaigns, leads, LinkedIn and email workflows.
Communication CRM Marketing
1011
SalesTouchOAuth
io.salestouch
Give AI agents the LinkedIn tools to find, qualify, engage, and follow up with prospects.
Social Media CRM Autonomous Agents
22
OmentirOAuth
io.github.vanshyadav1408
Find LinkedIn prospects, draft outreach, and run human-paced campaigns from your AI agent.
Social Media Marketing
144
Sliq AOAuth
com.getsliq
Run LinkedIn and email outbound from Claude, ChatGPT, or any AI agent.
App Automation Email Marketing
1591

Related MCP Servers

OutreachPilot MCP Server
CRM Communication
emielmadonna
F
license
A
quality
D
maintenance
Enables AI agents to automate sales outreach, research leads, and manage campaigns directly in OutreachPilot via natural language commands.
Updated 4 months ago (2026-05-28 20:55 UTC)
31
-
Outbound Engine MCP
Marketing Autonomous Agents
closermethod
A
license
A
quality
D
maintenance
Provides AI agents with tools for structuring and auditing cold outreach messages, building follow-up sequences, handling replies, and detecting red flags.
Updated 3 months ago (2026-07-14 14:06 UTC)
7
34 npm
MIT
LinkedNav MCP
CRM Social Media
linglistack
A
license
Not graded
quality
C
maintenance
Connects AI assistants to LinkedIn outreach, enabling lead finding, campaign management, messaging, and analytics through natural language.
Updated 2 months ago (2026-08-19 07:40 UTC)
MIT
Scrupp MCP serverofficial
Browser Automation Web Scraping
Wism-Corp
A
license
Not graded
quality
B
maintenance
Enables AI agents to export leads from LinkedIn, Sales Navigator, and Apollo searches, enrich profiles, look up companies, find decision makers, and discover or verify work emails without writing HTTP calls.
Updated 20 days ago (2026-09-15 12:11 UTC)
MIT