Bring your own coding agent

pm7Code, your ACEpm7Code, your Agentic Coding Environment.

A workspace where your projects live, your agent works, and your apps get built and shipped.

You and your coding agent of choice.

The full dev cycle, in one workspace

Projects
Agent
Build
Deploy

Six agents, one workspace

Pick the agent per session. Switch or hand a running session over to another one at any time — the same pm7Code session keeps its identity and transcript.

C
Claude Code
Anthropic, via the Claude Agent SDK
C
Claude CLI
Anthropic's Claude Code CLI
O
OpenAI Codex
OpenAI's Codex
P
Pi
Multi-provider, defaults to Google Gemini
C
Cursor
Cursor's agent over ACP
G
Grok
xAI's Grok over ACP

An optional Orchestrator mode lets one agent delegate sub-tasks to the others. pm7Code injects its own protocol into each agent through that agent's own system-prompt channel, so it never depends on a global config file to behave consistently.

One workspace. Zero context-switching.

Projects, agent, builds, deploys — all in the same place, all visible at the same time.

Bring your own agent

Claude Code, Claude CLI, OpenAI Codex, Pi, Cursor, and Grok. Hand a live session over to a different one without losing the transcript. No lock-in to any single coding agent.

Projects at the center

Your work is organized by project, not by which agent thread happened to run last. Sessions, prompts, and history live with the project.

Build & deploy in place

Run builds, ship deploys, watch logs without ever leaving the workspace. The thing you just told your agent to ship lands right next to the prompt that asked for it.

Command Center

Prompts, shelves, and history. Structured context you don't have to rebuild every time you start a new task.

CView transcript

A readable history of every agent run — questions, tool calls, results, summaries — that you can scroll back through, search, and replay.

Layered prompts

System, Group, Project, and Session prompts inherit from each other with a clear order. Lower layers override higher ones, and CView shows exactly what reached the agent.

Interactive questions

Agents ask you structured multiple-choice questions inline, with optional free-text. You answer in the workspace and the agent continues — no guessing, no context lost.

Neighbor-agent review

Turn on a second opinion per session: before a question, the active agent consults a designated buddy agent and shows both recommendations side by side.

Git Service

Agents hand off commit and push as a background job. The exact files are staged and pushed while the session keeps working — with a visible status card and strict, no-surprises rules.

Browser View

An optional in-app browser panel right of the transcript, with its own tabs. Preview the deploy you just shipped without leaving the workspace.

Live Summary

Independent summary slots distill a long run into the gist as it happens, each with its own provider and model — so you can scan what an agent is doing without reading every line.

Run agents remotely

Route a session through a remote websocket motor so the agent runs on another machine while the workspace, transcript, and status stay local and live.

Portable memory

pm7Code keeps its own memory of facts, preferences, and conventions — agent-independent, scoped per project or globally, and recalled automatically. The agent proposes; pm7Code decides.

Local-only

Runs on your Mac and drives the agent CLIs you already have. No telemetry to PM7, no cloud sync.

Independent by design

pm7Code carries its own protocol into every agent. It never depends on a global config file to behave the same way.

Efficiency, not novelty

Every panel earns its place by removing a step between you and the next build. Shipping faster, not collecting features.

Your projects. Your agent. One workspace.

pm7Code keeps everything between "I have an idea" and "it's shipped" in the same place.