Open Labs

Agentic coding environment

Keaz

An agentic coding environment built around control.

Keaz brings terminal-native agent work, editor context, permissioned changes, and reviewable sessions into one local workflow for serious repositories.

Keaz feature gap report dashboard
Terminal-native Interactive, one-shot, and terminal-native flows share the same engine.
Controlled edits Workspace policies, sandbox profiles, approvals, and reviewable diffs stay explicit.
Editor bridge VS Code, Cursor, and Windsurf can hand active-editor context to local sessions.
Session memory Resume, fork, inspect, and audit work instead of losing agent state.

How Keaz works

The terminal is the first-class IDE surface.

Keaz starts where developers already work: the repository, shell, and editor. The runtime can read project instructions, inspect files, stream model output, run approved tools, propose edits, and preserve the session trail for later review.

  1. OrientLoad workspace context, instructions, git state, and visible editor metadata.
  2. PlanKeep the user in control before writes, commands, or broader permissions happen.
  3. ExecuteRun shell, file, search, MCP, and editor bridge actions through one policy layer.
  4. ReviewOpen diff previews, keep or revert changes, and preserve the session evidence.

Product surface

Built for the practical work around agentic coding, not only the prompt.

Runtime

Runtime boundary

Rust agent core, provider adapters, tool registry, permission engine, protocol, and state store.

Workflow

Daily developer loop

Interactive CLI, TUI panes, `keaz exec`, slash commands, diagnostics, and resumable sessions.

Editor bridge

Editor control

Active-file context, diagnostics, visible tabs, changed-file previews, and bridge doctor reports.

Trust

Trust mechanics

Sandboxing, path grants, approval memory, protected metadata paths, and dry-run-first setup flows.

Why it matters

Agentic coding tools become process runtimes the moment they touch real repos.

Keaz treats process execution, file mutation, provider routing, editor context, and state recovery as core product architecture. That makes it a stronger foundation for coding agents than a UI-only assistant.

RustCore runtime and CLI workspace
SQLitePersistent sessions and auditability
MCPExternal tools behind permission policy

Product portfolio

Move across the Open T Labs product portfolio.

Next step

Bring the agentic coding workflow you want to make real.

Open T Labs can walk through Keaz as a product direction, implementation reference, or scoped agentic IDE build path.