← Builders / TUI

GA Universal coding harness Any model · switch any time

Coding agents went solo.
Your team didn't.

imm tui brings a universal coding harness to your terminal, real file edits, shell access, plan-first autonomy, any model you choose, running inside your team's zones, shared memory, and playbooks. The agent doesn't start from zero. It starts from what your team already knows.

$ Get imm tui See how it's different →

CONVERSATION PICKER · EVERY ZONE, EVERY CONVERSATION

In My Mind TUI conversation picker showing zones (MIL, Software Engineering, Unzoned, Massive Exec), teams with a Lead and members carrying roles like Tech Lead, Backend Engineer, DevOps Engineer, QA Engineer, model per member, and last-active columns

Filter by title, zone, or model. Every conversation in the workspace on one screen, solo chats and full team runs side by side.

SOLO · SINGLE MODEL

In My Mind TUI single-model session with grok-4.5-extended, agentic-mode banner surfaced, guarded posture with per-tool prompts, session timer and context usage in the status bar, keyboard hints along the footer

One model, one job. Guarded posture on. Agentic mode is opt-in per session, never assumed.

TEAM · LEAD + 4 MEMBERS

In My Mind TUI team conversation, Tech Lead running grok-4.6 with team mode just toggled from lead-only to 4 members ready to delegate, same status line and posture controls as the solo view

Same status bar. Same shortcuts. Flip on the team and the Lead delegates in-thread to the members you picked.

// why it's different

Your workspace's beliefs, brought to the terminal.

Five reasons a team coding environment beats a solo agent. Every one shipped.

MEMORY

Shared memory, not a fresh session

Architecture decisions, conventions, and prior work live in zone and workspace memory the whole team shares. The agent starts knowing what your team already figured out.

STANDARDS

Standards, encoded once

Review checklists, conventions, playbooks, and standing orders the agent follows every time. Consistency without policing.

SHARED PLANS

Plans and runs in shared zones

Plan-first by default. The agent proposes before it touches anything, and the plan and the run live in a shared zone your team can see, not a private terminal.

COUNCILS

Councils for the hard calls

Code-review and architecture councils put several models on a decision. A built-in second opinion, not one model's best guess.

GOVERNANCE

Governed by design

Per-zone permissions, approval postures, an always-on OS-level sandbox, and an audit trail. Autonomy a team can actually allow.

THE MOAT

Knowledge compounds. Context never dies.

Every session deepens the workspace. What one developer's agent learns, the team's workspace keeps, so tomorrow's run is smarter than today's.

// the how

One harness. Any model. Every surface.

The capability underneath. The same coding-agent loop no matter the model or the surface.

  • → Reads, edits, runs, and plans. Real file tools plus shell on real projects.
  • → Plan-first. Investigate and propose before touching anything.
  • → Approval on a dial. Read-only planning up to full trust, one keystroke.
  • → Always-on OS sandbox. Write-jailed to your working directory.
  • → You see it think. Live reasoning, a running to-do list, labeled diffs.
  • → Native binary in the terminal (imm tui), native desktop app, same workspace as the web.

MODEL POSTURE

Choose your model. Switch any time.

Start with the best frontier models. Add specialists as your work demands. Bring your own. Switch mid-conversation, the workspace, memory, and plan carry over.

Frontier models · Specialized models · Your models · One workspace

// the contrast

vs. the solo coding agent.

THE SOLO CODING AGENT

  • A single session with no memory of your team.
  • Your standards, re-explained every time.
  • One model, one answer, no cross-check.
  • A private terminal, nobody else sees the work.

IN MY MIND

  • Shared memory that compounds across the whole team.
  • Playbooks and standards the agent follows automatically.
  • Councils, plus any model you choose, switchable any time.
  • Shared zones, governed, audited, seen by the team.

// one workspace, every surface

Type in the terminal. Pick it up in the browser.

The TUI is a typed terminal surface. The web app is typed in the browser. The Desktop is typed in a native window. They share the same zone memory, same library, same tools, same access controls. Switch surfaces mid-conversation and the team picks up where you left it.

Voice agents run in the browser and on the phone (WebRTC), not in the terminal, but they hit the same workspace. A voice call's transcript lands in the same zone you can later open in the TUI to follow up.

Where Humans and AI collaborate.