Skip to main content

​
What is Shanone?

Shanone is not an autonomous agent that thinks and plans on its own. Shanone is the layer underneath the agents your team already uses — Cursor, Claude Code, or any other Model Context Protocol (MCP) client. It is:
  • A tool vendor — a single hosted MCP server that exposes Slack, GitHub, Notion, Stripe, Google Workspace, Ethereum, and 140+ other services as callable tools, so your agents don’t need a bespoke integration for every service they touch.
  • A skill vendor — a place to save reusable, step-by-step workflows (“Skills”) once and reuse them from any machine, IDE, or agent.
  • A permission manager — an IAM-style control plane so the team (or organization) that owns the Shanone account decides exactly which member can use which tool or service, not just who has an account.
Put simply: tools for your agents, permissions for your team, all in one place.
Shanone

​
The numbers

Shanone is marketed as offering 140+ integrations and 10,000+ tools — this is the full catalog Shanone is built toward across its integration ecosystem. As of this writing, 269+ service integrations are implemented and live in the Shanone backend (apps/backend/src/tools/), spanning categories like:

Communication

Slack, Discord, Telegram, Microsoft Teams, Google Chat, LINE, and more

Developer platforms

GitHub, GitLab, Bitbucket, Linear, Jira, CircleCI, Sentry, Vercel

Productivity & docs

Notion, Google Workspace (Gmail, Calendar, Docs, Sheets, Drive), Confluence

Business & finance

Stripe, HubSpot, Salesforce-adjacent CRMs, QuickBooks-adjacent tools, Airtable

Web3

Ethereum JSON-RPC, ENS, token/NFT data via Alchemy

AI & infra

OpenAI, Anthropic, Modal, Replicate, Datadog, Cloudflare, AWS-adjacent services
Don’t see a service you need listed on the Integrations pages? It’s very likely already implemented — those pages only cover a curated subset in depth. Ask your agent to run shanone_search_tools with the service name; if it’s registered, it will show up.

​
How agents use Shanone

External AI agents don’t get a separate MCP tool for every possible action (that would blow out their context window). Instead, Shanone exposes a small number of meta-tools that follow a search with schema → execute pattern, similar to how Composio structures its tool catalog:
1

Search for a tool

The agent calls shanone_search_tools with a natural-language query (e.g. “post a message to a slack channel”) to find the right tool name. For a full listing of everything a service offers, it uses shanone_list_service_tools instead.
2

Inspect the schema

The agent reads inputSchema in the search result. It calls shanone_get_tool_schema only when the result has schemaRef instead.
3

Execute the tool

The agent calls shanone_execute_tool with the tool name and arguments. If the underlying service needs authentication, Shanone returns a connect_link instead of failing outright — the user visits it, authorizes, and the agent retries.
The same pattern repeats for reusable workflows: shanone_list_skills → shanone_get_skill → the agent follows the instructions itself, calling tools as needed. And for team administrators, shanone_list_services → shanone_set_user_tool_permission / shanone_set_user_service_permission lets a Root or delegated SA2 user control exactly which tools each teammate’s agent is allowed to run. All of this happens over a single MCP connection — no per-service OAuth apps to register, no per-service SDKs to vendor into your agent’s runtime.

​
Next steps

Quickstart

Get an API key, connect your first client, and run your first tool call

MCP Overview

Understand the meta-tool workflow and how Shanone’s MCP server is structured

Integrations

Browse the integration catalog, including Slack, GitHub, Notion, and Stripe

Permissions & Access Control

Learn how teams control who can use which tools