Claude Code · Codex · Gemini CLI · OpenCode
It’s no longer just text — it’s a full interface.
Mirafold gives your work the experience it calls for.
Same agent, same machine: it keeps its existing settings, memory, permission rules, and credentials.
Free and open source · MIT · Node 22+
what changes
Your agent edits files, runs tests, tracks tasks, produces diffs, asks questions, and reports progress. In a terminal, all of it arrives as one stream of text. Mirafold gives each kind of work the interface it calls for — and leaves text where text is right.
A change is a diff you can inspect as a change. Progress is a live checklist, not prose buried in history. A comparison is a table, a measurement is a chart, an architecture is a diagram — each appears when it is the better interface for the information, generated by the agent as it works.
Continue from a card. Review the real difference between Git HEAD and the working tree in its own Workspace changes view, not in conversation text. Select a hunk, then put its exact code context into the prompt with Explain or Request change — still editable, not sent for you. Permission requests arrive as controls, in a strip the agent cannot touch.
Mission control puts all your running agents in one grid: their state visible, actions one click away, no hunting through terminal tabs.
Thinking, tool calls, diffs, subagents, task status, tokens, and
cost all stay available — the full interface is a superset. Prefix a command with ! for a real
PTY where interactive sudo and ssh work.
Pro brings the same session to a phone, tablet, or another browser. The end-to-end-encrypted relay sees ciphertext and connection metadata, never your key, code, or content.
Pin a rendered block and it stays visible, updating as the work changes. When the built-in forms are not enough, the agent can ship arbitrary UI into a no-network, opaque-origin sandbox.
$0 open source, MIT
$12/mo · $99/yr
The Mirafold experience, from any device.
Yes — a browser interface for the terminal coding agent you already
use, rendered faithfully: same engine, same settings, memory, and
permission rules as your terminal. Install with
npm i -g mirafold, run mirafold in a project
directory, and your agent opens in the browser with a full interface for
its work — diffs you can inspect, live task lists, tables, charts, and
pinned dashboards — instead of one scrolling stream of text.
Yes. A session is fully local when the agent behind it points at a local inference server — Claude Code against Ollama’s Anthropic-style endpoint; Codex against Ollama, LM Studio, or vLLM; or OpenCode through a local provider in its own config. Then nothing — code, key, or tokens — leaves your machine. The repo documents the exact recipes and an honest model/hardware table.
In your local daemon’s environment, exactly as in a terminal —
ANTHROPIC_API_KEY and friends. It is never serialized to
the browser, and on Pro the relay carries end-to-end-encrypted
frames plus the connection metadata any forwarder handles (IPs, timing,
byte counts) — never the key or the conversation.
Claude and Gemini subscriptions are refused because their terms prohibit third-party use; API keys are the path for those agents. A Codex/ChatGPT sign-in works locally with the caveat shown at sign-in: OpenAI hasn’t clearly permitted third-party use in writing, but visibly tolerates it — your account, your call. OpenCode can use provider API keys or local providers. Its built-in Zen gateway is available in local-only sessions; free-period prompts may be used to train models. Its ChatGPT route carries the same caveat, and other subscription logins are not usable. No subscription login or Zen session runs over the relay.
Two zones, hard boundary. The agent’s output renders only into an output zone: streamed markdown (no raw HTML), schema-validated registry components, and — for arbitrary UI — a sandboxed iframe with an opaque origin and a CSP that cuts every network path. The prompt box, socket, and credentials live in a trusted shell the agent cannot render into, and anything a component does back is validated server-side against an allowlist.
Claude Code, Codex, Gemini CLI, and OpenCode ship today, each behind its own adapter — none privileged. A new agent is one adapter behind a small interface, not a rewrite; the adapter spec in the repo documents exactly what an engine needs to qualify.
Yes — run as many sessions as you want, each with its own directory and agent. Mission control shows every session in one grid, at a glance, and you can act on any of them straight from the grid — no hunting through terminal tabs or tmux panes.
They work, in the same page. Type ! before any command
and it runs in a real PTY on your machine, so interactive prompts
behave: sudo’s password prompt appears in the shell’s own masked
input — the agent never sees what you type — and afterward the agent
reads the command’s output and picks up from there.
Yes — that’s Pro: pair a phone, tablet, or any browser with a QR code, and the same session travels with you — full interface, pinned widgets and all. The relay in between is end-to-end encrypted — it sees ciphertext and bare connection metadata, never your key, your code, or your content. The relay’s code is open source (MIT) like everything else; the hosted relay is the convenience Pro pays for.
The whole product is MIT — shell, daemon, all four adapters, and the relay code itself, so you can read exactly what the piece you’d be trusting can and cannot see. Pro is convenience infrastructure: the hosted relay that lets your devices reach your sessions from anywhere, end-to-end encrypted. If you’d rather self-host the relay, you run the same code we do.
Your agent, your key, your machine — and a full interface for the work.
