---
title: "Toolsets"
description: "Which of the 123 tools a connection sees, and how to narrow them."
---

> Documentation Index
> Fetch the complete documentation index at: https://docs.heizen.work/llms.txt
> Use this file to discover all available pages before exploring further.

# Toolsets

A connection only sees the toolsets it asks for. `all` — what the
[setup examples](/mcp/clients) use — turns on every one of them. `core` is
always enabled even if you ask for nothing; anything else comes from the
`x-heizen-toolsets` header, either `all` or a comma-separated list.

| Toolset | Tools | Contents |
|---|---|---|
| `core` | 31 | Projects, bugs, stories, epics, sprints |
| `tests` | 27 | Test plans, suites, cases, environments, executions, runs |
| `docs` | 27 | Wiki, folders, resources, requirements, scopes, personas |
| `delivery` | 8 | Worklogs, story comments |
| `meetings` | 14 | Recordings, notes, transcripts, calendar, bot control |
| `integrations` | 15 | Jira, Confluence, GitHub, Figma, PR linking |
| `api` | 1 | `studio_api_call` escape hatch for endpoints without a tool |

Unknown toolset names are ignored rather than rejected. The gate applies to
`tools/call` as well as `tools/list`, so a hidden tool can't be invoked by
guessing its name. Call `list_toolsets` in a session to see what's active.

> **When to narrow it**
>
> 123 tools is a lot of context, and a long tool list makes some models pick
> worse. If you notice that, swap `all` for the sets you actually use — `core`
> alone covers backlog and bug work, `core,tests` covers QA, add `integrations`
> when you're pushing to Jira. Clients that can't send headers are stuck on
> `core`; bridge through `mcp-remote` when you need more.

For what each set actually contains, see [Tools](/mcp/tools).

Source: https://docs.heizen.work/mcp/toolsets/index.mdx