Documentation

Tools

Give your AI the ability to take action — integrations like Gmail, Outlook, Slack and Calendar, your own HTTP APIs, and functions you write yourself.

Tools are the actions your assistant can take on your behalf. Where resources let it read your knowledge, tools let it do things — send an email, post to Slack, book a meeting, search the live web, or call your own API. Every project ships with a set of built-in tools, and you add more whenever you need them.

Three ways to add one

The Tools page has three tabs, matching the three things you can put on your server:

Tab	What it is	Plan
Functions	Code you write. Multi-step logic, transforms, anything that combines a credential with a computation.	Pro
HTTP Endpoints	One request against an API you already run, described in a form. No code.	All plans
Catalog	The integrations we ship and maintain — connect an account, switch on the tools you want.	All plans

Start at the catalog. If your case is covered there, connecting an account is the whole job. If it isn’t, an HTTP endpoint is the shortest path to your own API, and a function is what you reach for when one request isn’t enough.

Why use tools

Go from answering to acting. Your AI doesn’t just reply — it books the meeting, sends the follow-up, and files the ticket.
Connect once, reuse everywhere. Authorize an integration a single time and every enabled tool works across all your linked channels and MCP clients.
Enable only what you need. Turn tools on one by one, so the assistant can do exactly what you allow — and nothing more.
Bring your own API. Add a custom HTTP endpoint or connect a vendor’s remote MCP server to expose services we don’t ship out of the box.

How a tool works

Under the hood, your project is an MCP server — and a tool is just a function that server exposes to any connected AI. The flow is always the same:

You enable a tool. It’s added to your assistant’s toolset and described to the model — its name, what it does, and the inputs it expects.
You connect the integration once. Tools that touch an outside account (Gmail, Slack, Calendar…) need access. You authorize the integration a single time — with OAuth or an API key — and Ganju stores the credential securely and refreshes it automatically. Built-in tools need no connection.
The model decides to call it. When someone makes a request in a channel or client, the model reads the available tools and, if one fits, calls it with the right arguments — no code from you.
Ganju runs it and returns the result. The call runs with your stored credentials, scoped to only the permissions that tool needs, and the result flows back to the model to finish the reply or complete the action. Every call is recorded toward your usage.

Default tools

Every project starts with five built-in tools already installed, so your assistant can work with your resources and prompts from day one — no connection required:

List Resources — list every resource available to this assistant.
Read Resource — read the contents of a stored resource.
Send Resource — deliver a resource to the user as a chat attachment.
Search Resources — find the resources most relevant to a question using semantic search.
List Prompts — list the prompts and commands this assistant exposes, and how to run them on the current channel.

Turning tools on and off

Every tool has a switch and a delete, and they do different things. Off stops the tool being exposed while keeping its configuration and its connection; Remove deletes the row and takes the settings with it. Turning a tool off frees a slot against your plan’s tool count, so you can rotate through more than you expose at once.

Keep the list short on purpose. Every enabled tool’s schema is re-sent to the model on every call, so a long tool list costs tokens on every turn and makes the model’s choice harder. Channels cap the list at 40 for that reason.

Available tools

Each integration has its own page covering what it does, how to connect it, and every tool it offers:

Built-in — the five core resource and prompt tools every project ships with.
Gmail — send, read, search, and manage email (18 tools).
Outlook — Microsoft 365 mail via Graph (18 tools).
Slack — post messages, browse channels, and upload files.
Slack Search — workspace-wide message search (separate user-token connection).
Google Calendar — create and manage events and find open time slots.
Cal.com — check availability and book or cancel meetings.
Web Search — search the live web and extract page content, powered by Tavily.
GitHub — connect GitHub’s official remote MCP server for repos, issues, and pull requests.
Notion — connect Notion’s official remote MCP server to search and update pages and databases.
Greeting — a tiny demo tool for testing a new server.

And the two you build yourself:

HTTP Endpoints — expose your own APIs as named tools, with no code.
Functions — write your own tools in JavaScript, from the browser or the ganju CLI.

Next: decide where people use your assistant — set up channels.

Last updated September 4, 2026