Code — overview

Ship software in Miton: chat-first agent, editor, code search, sub-agents, per-turn rewind, and project intelligence. Overview and links to task guides.

Product stage
Development build
Last reviewed
2026-08-27
Successful result
The change is visible in the active project.

What this accomplishes

Ship software in Miton: chat-first agent, editor, code search, sub-agents, per-turn rewind, and project intelligence. Overview and links to task guides.

Code is Miton’s software engineering lens. You get a chat-first agent that knows your project, a code editor, search, sub-agents in isolated worktrees, per-turn rewind, and project intelligence for milestones and drift.

This page is the overview. Task guides below cover each tool. See also /for-developers.

Current availability

This page documents behaviour verified in Miton development builds. Miton is still pre-release and is not publicly downloadable; invited beta users will receive a signed desktop build with automatic updates.

Before you start

  • A Miton development build and an active project containing source code.
  • A connected model. Trust the project before allowing file, terminal or Git actions.

What Code does

Code is for engineers who ship real software. You can:

  • Open a project and let the agent read structure before you type.
  • Run tests in a sub-agent — ask “is the build green?” and get real results.
  • Review diffs in a sub-agent with structured feedback.
  • Spawn sub-agents in isolated git worktrees parallel to your main session.
  • Rewind any agent turn with one click — git-native restore points.
  • Post inline PR comments on GitHub, GitLab, and Bitbucket with sensitive content stripped locally.
  • Track milestones, blueprints, and drift between plan and codebase.

How Code fits with the other lenses

Miton shares one agent engine across Chat, Writer, Canvas, Code, and Automation:

  • Writer — long-form editor, citations, export.
  • Canvas — visual canvas, design tokens, forms.
  • Automation — workflow graph, scheduler, webhooks.

They share memory, model routing, permissions, and undo.

Default tools

Code highlights these agent tools:

  • read / write / edit — read and write files with diff preview.
  • bash — shell commands with permission prompts.
  • glob / grep — file and content search.
  • pr — open a pull request on GitHub, GitLab, or Bitbucket.
  • subagent — spawn one of the named built-in sub-agents.
  • specialist_team — propose a temporary team of task-specific Specialists for review.

You can disable, configure permissions, or add custom tools.

Built-in sub-agents

Eight default types:

  • Explorer — read-only codebase exploration.
  • Verifier — run tests and report results (read-only).
  • Debugger — investigate failures; can add instrumentation.
  • Test Runner — continuous test runs with failures surfaced in chat.
  • Orchestrator — coordinate multiple sub-agents for multi-step plans.
  • Researcher — cited web research, one subtopic per parallel run.
  • Designer — critiques and improves visual designs, applies Canvas patches.
  • Analyst — loads and analyses data files, produces charts and findings.

Spawn any of them from chat with @<name> (for example @explorer, @researcher). For an objective that doesn’t fit a built-in type, the agent can propose a temporary Specialist team instead — see Sub-agents.

Pages in this section

Shared topics live in the Docs index. Engineering detail lives in ../../architecture/.

What success looks like

Ship software in Miton: chat-first agent, editor, code search, sub-agents, per-turn rewind, and project intelligence. Overview and links to task guides. The result remains attached to the active project when you move to another workspace.

Common failures and recovery

  • Confirm the correct project and branch are active.
  • Check the model connection and the permission request shown in Chat.
  • Keep the exact error before retrying; use Code troubleshooting when the failure persists.

Open Code troubleshooting for recovery steps and diagnostic paths.