Docs

Skills and Connectors

MindMux splits two extension concepts into separate product surfaces: Skills and Connectors. Both are managed under Settings, but they have very different responsibilities.

Skills

A Skill is a scenario-specific Markdown rulebook that shapes how the agent behaves in a given situation.

Key qualities

  • Markdown-defined: each skill is a Markdown file with frontmatter (name, description, when to use) plus a body of detailed rules.
  • Three scopes:
    • Built-in: shipped with the app, available in every workspace, and can be individually enabled/disabled by the user.
    • Global: user-defined skills shared across workspaces.
    • Workspace: stored in brain/skills/, effective only for the current project.
  • Lazy loading: at session start only the list of available skills is injected into the system prompt; the full rules are loaded on demand via invoke_skill when the agent decides they are relevant.
  • Built-in enable/disable: users can turn off unwanted built-in skills in Settings > Skills to keep the system prompt clean.

Boundary with Pages

SkillPage / Root Doc
AnswersWhat the AI should do in a given situationWhat the project knows
Locationbrain/skills/ (workspace skills)brain/pages/
NatureAI operating procedureProject knowledge

Do not write project design docs as skills, and do not write AI execution rules as pages.

Connectors

Connectors are the MCP connector layer that brings external systems into MindMux, letting the agent call tools exposed by those systems.

Current scope

  • Official connectors: GitHub, Notion, Linear, GitLab.
    • GitHub / Linear: hosted HTTP with bearer token.
    • Notion / GitLab: OAuth-first connectors.
  • Custom MCP V1: supports stdio, legacy SSE (via a local stdio bridge), and streamable HTTP; auth modes include none, bearer, api key header, and custom headers.

Management entry points

  • Settings > Connectors: manually add, enable, disable, test, and authorize connectors.
  • Configuration expert: a conversational entry point that shares the same connector state machine with the Connectors UI.

Why they are separate

  • Skills change how MindMux behaves (how the AI thinks and distills knowledge).
  • Connectors change which external systems MindMux can talk to.

Merging them would blur the concept of an "extension". Keeping them separate makes both the architecture and the settings UI clearer.

Settings entry points

The Settings sidebar is a flat list of 6 items:

  1. General
  2. Runtime (provider profiles)
  3. Experts
  4. Connectors
  5. Skills
  6. Workspace