Provides an OAuth connector for Atlassian products (e.g., Jira, Confluence), enabling agents to interact with Atlassian APIs.
Offers a Bearer/PAT connector for GitHub's MCP server, allowing agents to access GitHub repositories and operations.
Integrates with Linear via OAuth, allowing agents to manage issues, projects, and teams programmatically.
Connects to Notion via OAuth, enabling agents to read and modify pages, databases, and workspace content.
Provides an OAuth connector for Sentry, enabling agents to access error tracking, issues, and monitoring data.
Integrates with Stripe via OAuth, allowing agents to interact with payments, customers, and billing APIs.
Connects to Supabase via OAuth, enabling agents to work with databases, authentication, and storage services.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@scriptit set up a watch that runs list_cronjobs every 5 minutes and alerts me if any fail"
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.
One control plane for your agent's tools. Connect remote MCP servers once; every agent — in any sandbox — gets them back two ways:
a remote MCP endpoint (/mcp) with compact meta-tools (no tool-list bloat), and
a CLI (scriptit) an agent can pair from inside any sandbox via a public-key
handshake approved over an MCP tool call — upstream credentials never enter the sandbox.
Plus a proxied LLM endpoint for scripting, and watches: run a command or tool on an interval, optionally have an LLM verify the output, and emit structured events your agent can poll or stream.
Local-first build: everything runs on your machine (SQLite, one server process).
pnpm install
pnpm build
pnpm start # control plane on http://127.0.0.1:4747 (prints admin + mcp tokens)Dev mode (server + web UI + demo upstream MCP, all watching):
pnpm devDashboard: http://127.0.0.1:4747 (dev UI: http://127.0.0.1:5173) — sign in with the admin token.
Demo upstream MCP server: http://127.0.0.1:4748/mcp (echo, get_time, list_cronjobs).
LLM features need ANTHROPIC_API_KEY (or an ant auth login profile) in the server's env.
Related MCP server: ChatGPT Gateway MCP Nyan
Add to Claude Code (or any MCP client):
claude mcp add --transport http scriptit http://127.0.0.1:4747/mcp --header "Authorization: Bearer <mcp token>"Tools: list_connectors, search_tools, call_tool, approve_cli, create_watch,
list_watches, run_watch, delete_watch, get_events, llm.
In the sandbox / any shell:
scriptit pair # prints a code like AB3D-9XKQ and waitsThen either ask your agent to call approve_cli with that code over its ScriptIt MCP
connection, or approve it in the dashboard. From then on:
scriptit connectors
scriptit call demo list_cronjobs
scriptit llm "summarize:" - < notes.txt
scriptit watch create --name failed-crons \
--connector demo --tool list_cronjobs --interval 5m \
--verify "Emit an event iff any cronjob is failing; type=cron_failure, summarize which."
scriptit events --follow --json # structured events, one JSON per lineThe CLI signs every request with an ed25519 key; sessions are short-lived and revocable from the dashboard. Upstream tokens live only in the control plane, encrypted at rest.
Path | What |
| control plane: REST + MCP endpoint + OAuth broker + watch engine (port 4747) |
| dashboard SPA |
|
|
| shared types + ed25519 signing |
| demo upstream MCP server (port 4748) |
| the contract: pairing, signing, REST, MCP tools, watches |
Local build trusts localhost: admin/mcp tokens are the auth boundary. Don't expose 4747.
OAuth connectors use Dynamic Client Registration + PKCE (works with Linear, Notion, Sentry, Atlassian, Stripe, Supabase, …). Bearer/PAT connectors cover GitHub's MCP and friends.
scriptit watch command-type watches execute shell commands on the server host — that's
the point (local build), but remember it when you deploy this anywhere shared.
This server cannot be deployed
Agent-native collaboration network: orchestrate a team of long-running agents from any MCP client.
MCP-first toolbox for agents: KV storage, auth, queue, and utility tools. Free in early access.
The Remote MCP server acts as a standardized bridge between LLM applications (like Claude, ChatGPT, and Cursor) and external services, enabling AI agents to access external tools and resources. Its primary capability is providing a centralized search tool to discover other MCP servers and their respective tools. Unlike local implementations, it runs remotely with OAuth authentication and permission controls for security.
Remote MCP server to run your Atako AI agents: chat, projects, files, integrations and channels.