Log inSubmit server
ProAgentStore logo

ProAgentStore

Manage ProAgentStore agents, private instances, knowledge, and runtime chats through MCP.

InstallRepositoryWebsite
Full README Compare alternatives Embed widget Claim this listing

Install ProAgentStore

Pick your client, then add it in one click or paste the config.

STDIONode.js
~/Library/Application Support/Claude/claude_desktop_config.json
claude_desktop_config.json
{
  "mcpServers": {
    "proagentstore": {
      "command": "npx",
      "args": [
        "-y",
        "@proagentstore/cli"
      ]
    }
  }
}

Settings → Developer → Edit Config, merge this into mcpServers, then fully quit and reopen Claude Desktop.

Install not yet sandbox-tested— we haven't run this command through our automated sandbox yet. Not a red flag; we're working through the catalog.

Overview

ProAgentStore MCP server connects MCP clients to the ProAgentStore marketplace, public agent trials, and private agent instances. Its tools support agent discovery, subscriptions, instance management, knowledge updates, and conversations with either public or user-owned runtimes. Use the hosted streamable HTTP endpoint for discovery and testing, or for operating subscribed agents with their own state and AI credentials. Server-side scopes, confirmations, dry runs, read-only mode, and audit events govern mutating and runtime actions.

Use cases

  • Discover and trial published ProAgentStore agents
  • Subscribe to an agent and manage its private instance
  • Add knowledge before chatting with an instance
  • Run browser-capable agents from a local machine
  • Audit MCP mutations and runtime activity

Key features

  • Public agent discovery and trial chats
  • Private instance subscription and conversations
  • Instance knowledge management
  • OAuth scopes and read-only enforcement
  • Dry-run and confirmation safeguards
  • MCP audit event access

Security & health

What it can reach and how reliably it runs — from our own checks, not the maintainer’s claims.

ActiveLast checked 1d ago

Recent health check succeeded.

19/21 checks healthy over the last 90d

Supply-chain signal

No high-severity advisories surfaced by our automated scan.

Critical 0High 0Medium 0Low 0

Scanned 1d ago via OSV.dev · @proagentstore/cli (npm)

Adoption & maintenance

Factual signals from GitHub, npm, and our automated checks — not a rating.

GitHub stars1Stargazers on the source repository.
npm downloads24kPackage downloads in the last 30 days.
Last commit1d agoMost recent push to the default branch.

How ProAgentStore works

What ProAgentStore MCP server does

ProAgentStore exposes the ProAgentStore marketplace and agent control plane through Model Context Protocol. MCP clients can discover published agents, try a public agent without signing up, subscribe to an agent, inspect owned instances, add knowledge, and send messages to a private instance.

The platform supports three agent forms: conversational workers built from the worker template, scheduled workers built from cron, and stateless API tools built from api. A subscribed instance runs with its own state and knowledge. The instance runtime uses caller-provided Cloudflare Workers AI credentials rather than charging the platform's AI account.

How it works

The hosted MCP endpoint is available at https://mcp.proagentstore.online/mcp. A typical workflow starts with list_agents, continues with subscribe_agent, then uses my_instances, add_instance_knowledge, and chat_with_instance. For quick discovery or smoke tests, chat_with_agent calls a public trial endpoint instead of creating a private instance.

Private runtime calls require the instance owner's Cloudflare Workers AI account ID and API token. If those credentials are absent, the runtime returns an instruction to add them before running the agent. This indicates that the private execution path is active while refusing to use the platform's credentials.

Browser- and coding-capable instances can also use the ProAgentStore local browser runtime. The pags up command serves active runtime-capable instances through an outbound WebSocket relay, with Playwright and terminal or coding capabilities available locally. Cloud-only agents without a runtime capability do not require that process.

Setup and configuration

Connect an MCP client to the hosted endpoint using its streamable HTTP transport. Examples include:

  • Codex: codex mcp add proagentstore --url https://mcp.proagentstore.online/mcp
  • Claude Code: claude mcp add --transport http proagentstore https://mcp.proagentstore.online/mcp
  • Generic proxy: npx mcp-remote https://mcp.proagentstore.online/mcp
  • Local CLI proxy: npx @proagentstore/cli mcp

Codex and Claude Code may require an MCP login flow when the server reports that the client is not authenticated. The hosted service also provides OAuth scopes for read, write, runtime, and destructive operations. Set MCP_READ_ONLY=1 to force read-only behavior where that configuration is available.

For local browser or coding execution, install the CLI with npm i -g @proagentstore/cli, sign in with pags login, and start the runner with pags up. Headless and device-based sign-in options are available for machines without a display.

Tools and capabilities

  • Discover published agents and call public trial chats.
  • Subscribe to agents and list the caller's private instances.
  • Add knowledge to an instance and chat with its runtime.
  • Apply read, write, runtime, and destructive authorization scopes.
  • Use dry-run options and exact confirmation values for applicable mutations.
  • Review recent writes, runtime calls, dry runs, denied actions, and destructive events through mcp_audit_log.

Limitations and notes

Private instance execution depends on user-supplied Cloudflare Workers AI credentials. The local runner is needed only for instances that declare runtime capability; cloud-only chat, retrieval, and connector agents are skipped by pags up. The runner connects outbound through a per-instance WebSocket relay and does not require an inbound port or tunnel. Live instance membership is polled every 20 seconds when using the default watched runner mode.

Read the full README View source on GitHub

Reviews

No reviews yet — be the first to share how it worked for you.

Frequently Asked Questions about ProAgentStore

ProAgentStore is an MCP server that connects AI clients to the ProAgentStore marketplace and account APIs. Its main tools discover agents, run public trials, subscribe to agents, manage private instances and knowledge, and chat with subscribed runtimes using caller-provided AI credentials.

Connect your MCP client to `https://mcp.proagentstore.online/mcp` over streamable HTTP. Codex and Claude Code provide direct add commands, while `npx mcp-remote` and `npx @proagentstore/cli mcp` can act as alternatives.

Subscribe to the agent, find the resulting instance, add any required knowledge, and call `chat_with_instance`. The instance also needs your Cloudflare Workers AI account ID and API token before private runtime execution can start.

Yes. Use `chat_with_agent` to call the public trial endpoint, which is intended for discovery and smoke tests and does not create a private instance.

Install `@proagentstore/cli`, sign in with `pags login`, and start `pags up`. The runner handles active runtime-capable instances through an outbound WebSocket relay and supports headless or device-based sign-in.

Related MCP servers

All in Developer Tools All alternatives
Add a badge to your READMEShow it’s listed — live SVG, Markdown or HTML.

AllMCPs Directory Badge

Full Badge Customizer

Showcase your server listing on GitHub or your project documentation. Embed this dynamic SVG badge to highlight official listing status and live engagement.

Badge Style:
Live Dynamic SVG PreviewProAgentStore AllMCPs Directory Badge
Markdown (GitHub README)
[![AllMCPs](https://allmcps.com/api/badge/proagentstore?style=directory)](https://allmcps.com/mcp/proagentstore)
HTML Embed
<a href="https://allmcps.com/mcp/proagentstore"><img src="https://allmcps.com/api/badge/proagentstore?style=directory" alt="ProAgentStore on AllMCPs" /></a>