$ mayajaal connect

One terminal.
Every AI coding agent you run.

Mayajaal is the control plane for developers running multiple AI coding agents at once. Track Claude Code and GitHub Copilot sessions across every repo — live status, full transcripts, and where the day actually went, in a single pane of glass. Offline-first, so the terminal never waits on the network.

free while in early access · private beta · works with your existing CLI agents
mayajaal — session monitor — zsh
agenttaskstatusupdated
claude coderefactor auth middleware → jwt rotationRUNNING2s ago
copilotgenerate unit tests for payments/webhook.tsWAITING41s ago
claude codemigrate api routes to edge runtimeIDLE4m ago
copilotfix flaky e2e test: checkout.spec.tsRUNNINGjust now
claude codewrite changelog for v2.3.0 releaseIDLE12m ago

plugs into the agents you already use

Claude Code GitHub Copilot CLI Codex CLIsoon Cursorsoon Aidersoon

Built for devs drowning in agent tabs

Stop tab-switching between five terminals. Mayajaal gives every session a home, a status, and an audit trail.

Unified session feed

Every Claude Code and Copilot session streamed into one live, scrollable terminal feed — filterable by repo, agent, or status.

Real-time status

See which sessions are running, which are waiting on you, and which have gone idle — no more guessing which agent is stuck and which one is quietly blocked on a prompt.

Offline by default

The local database is the source of truth and the terminal dashboard never touches the network. Sync is a queue that drains when you reconnect — a flight or an outage costs you a delay, not your history.

Secrets stripped before disk

Agent transcripts pick up API keys, tokens and .env contents as a matter of course. Mayajaal redacts them on the way into the local database — not on the way out to the network — and marks every event it touched so you can see it happened.

Multi-repo, multi-branch

Track sessions across every repo and branch your agents are working in, with the git remote and branch captured per session.

Where the day actually went

One command breaks the day down by tool: how long Bash really took, how many calls it was, what it cost in tokens, which PRs shipped, and what is still moving.

Full session transcript

Every prompt, response and tool call recorded against the session, repo and branch it came from — so you can go back and see exactly what an agent did.

Three commands to full visibility

Mayajaal wraps your existing agent CLIs — no migration, no new workflow.

1

Install the daemon

A lightweight background process that watches agent CLI activity on your machine. Private beta today — with repo access it is one command from a clone to a binary on your PATH.

make install
2

Connect your agents

Registers hooks with Claude Code and Copilot CLI for live capture. No arguments — it finds what you have installed and leaves anyone else's hooks alone.

mayajaal connect
3

Open the dashboard

Watch every session live in your terminal. Sign in only if you also want it in the browser — nothing leaves the machine until you do.

mayajaal dashboard
# private beta — binaries are not published yet
# with repo access, from a clone:
cd mayajaal/daemon && make install

Common questions

Which coding agents does Mayajaal support?

Claude Code and GitHub Copilot CLI are what we're building against today — both verified against the CLIs' actual behaviour, not just their docs. Codex CLI, Cursor and Aider are next, through the same plugin interface.

What actually leaves my machine?

Nothing, until you run `mayajaal login`. Capture, storage and the terminal dashboard all work with no account at all. Once you do connect a device, sessions sync to your account so the web dashboard matches your terminal — and that includes prompts and assistant responses, minus anything the redactor stripped. `mayajaal logout` stops it and leaves your local history untouched. Everything works fully offline either way, because the local database is the source of truth and the cloud is a copy of it.

What about secrets in my transcripts?

Agent sessions pick up API keys, tokens and .env contents as a matter of course, so Mayajaal strips them before anything is written to disk — not just before it syncs. Files like .env, .pem and credentials are redacted wholesale rather than scanned. Redactions are marked in both the terminal and the web dashboard, so you can see one happened and count them per session. Pattern matching is never complete, though: treat it as one layer, not a guarantee.

Does it work offline?

Completely. The daemon writes to a local database and the terminal dashboard reads only from that — it never touches the network. Sync is a queue that drains when you reconnect, so a flight, a VPN, or an outage on our side costs you nothing but a delay.

Is there a CLI-only mode, no browser required?

Yes. mayajaal dashboard --watch runs a full-screen live TUI in your terminal. The web dashboard is optional, and so is signing in at all.

What platforms are supported?

macOS and Linux, as a single native binary with no runtime dependencies. Windows is planned but not yet tested.

Stop guessing what your agents are doing.

Install Mayajaal and get a live view of every AI coding session in under two minutes.