Terminal ticketing · Atlas Home · Built for AI agents

Keep the work in the repo.

Ask your coding agent for a project update. Atlas keeps tickets, blockers, and review history in your repository, with a table in the terminal and Kanban in the browser.

Install. Then ask.

Install Atlas once. Let your coding agent set up the project, then ask about the work in your own words.

  1. 01

    Install once

    curl -fsSL https://raw.githubusercontent.com/myrrazor/atlas-tasker/main/scripts/install.sh | sh

    One small CLI. Needs curl, tar, and gh. Checksum plus a local attestation bundle; no GitHub login.

  2. 02

    Ask your agent to set it up

    In the project you want tracked, ask your coding agent:

    Initialize Atlas Tasker in this project.

    That is tracker init: board, local checkpoints, and integrations for agents already on this machine. No picker. Restart the agent. Grok also needs you to trust this project before local skills appear.

  3. 03

    Ask for status

    What's the current status of this project?

    Get current tickets, the work in progress, and anything that needs your attention. Your agent reads the same board you do.

The installer places the binary only. Run tracker init in each workspace, then restart detected coding agents so they load Atlas MCP. A written config is not a live connection. Open tracker to create a board or manage tickets in the browser.

Grok Build actual session · v1.15 source
Recorded in Grok Build 4.6 (xhigh). Real session, synthetic tickets, v1.15 source build. Silent; pauses shortened.

Ask for status, add a high-priority task, and see it on the refreshed board. Grok reads eight tickets through Atlas MCP, creates APP-9, then confirms all nine. Download video · Read the transcript.

Codex, Claude Code, Cursor, and Grok Bot all on one board.

Assign each coding agent its own tickets. They read the same Markdown board, claim with leases, and hand work to a reviewer without a second tracker.

Shared board Grok Build · Cursor · Grok Bot
Synthetic Example App board. Three agents, one project, local v1.15.0. Silent.

Grok Build owns the upgrade test, Cursor drafts the MCP guide, and Grok Bot reviews the retry fix on the same board. Download video · Read the transcript.

tracker web serve assignees visible
Synthetic Atlas Tasker web board with tickets assigned to agent:grok-build, agent:cursor, and agent:grok-bot
Each card names its assignee. Grok Build, Cursor, and Grok Bot share the columns.
tracker board terminal table
Terminal board table showing assignee column for Grok Build, Cursor, and Grok Bot
The terminal table shows the same assignees and statuses as the browser board.

A board in your browser. A board in your terminal.

Same tickets, same rules, same audit trail. Use whichever front end fits the moment and switch whenever you like.

tracker web serve 127.0.0.1 · local only
Atlas Tasker v1.16 in dark mode, showing example tickets and agent assignments across the Kanban columns
A local Kanban board with six workflow columns and a separate Canceled disclosure. Synthetic review data.
tracker tui local · synced
Synthetic local review capture of the Atlas Tasker TUI board showing aligned ticket IDs, statuses, priorities, and titles
Terminal TUI board from a synthetic local review workspace. The default is a polished table; --style kanban selects optional cards.
tracker agent available builder-1 agent queue
Atlas Tasker terminal agent queue showing ready and blocked tickets with reasons
Each agent gets a queue that explains what is ready and what is blocked, with the exact next commands.

And a third surface for your AI agents. A stable JSON CLI and local stdio MCP server let agents inspect queues, work tickets, attach evidence, and hand off context.

Connect your agent

From ticket to evidence, nothing disappears into chat.

  1. Ticket APP-42

    Define the work and its dependencies.

  2. Claim agent:builder-1

    A lease keeps another agent from taking it.

  3. Prove run evidence

    Attach test output, diffs, logs, and screenshots.

  4. Gate review required

    Completion waits for the approvals you chose.

The ticketing system your AI agents can actually operate.

Built for AI agents

Queues, gates, and typed tools agents respect.

Agents see exactly what is ready and why the rest is blocked, with stable reason codes instead of guesswork. Tracked writes record an actor, reasons are required for tracked MCP writes and protected operations, review gates guard completion, and the MCP server exposes typed tools behind escalating permission profiles.

Meet the agent surface
Atlas Tasker agent queue output listing ready work and blocked tickets with reason codes
tracker agent available answers with tickets, reasons, and suggested next commands.

Lightweight by design

Markdown is the record. Everything else is disposable.

Tickets live under projects/ as plain files you can open in any editor. Changes land in an append-only event log, a small local index keeps queries instant and rebuilds itself from the files if it ever goes missing, and tracker doctor --repair rebuilds it whenever you want.

├── projects/APP/tickets/APP-42.md
├── .tracker/events/ append-only
└── .tracker/index.sqlite disposable

CLI + TUI + web board + MCP

Pick the view. Keep the same tracker.

Script with the CLI, work keyboard-first in the TUI, or open Atlas Home with tracker. Home is one loopback service on 127.0.0.1:7432. There is no hosted mode and no login system. One binary does all of it.

Tour the CLI
CLItracker board --project APP
TUItracker tui
HOMEtracker
MCPtracker mcp serve --global --tool-profile workflow

One binary. Then initialize here.

The one-line installer is the normal path. It fetches the latest published release, checks its SHA-256 checksum, verifies the GitHub build attestation, and installs tracker. Workspace initialization is a separate, explicit step. After install, run tracker init in each workspace and restart those agents. See verification results on the GitHub release page.

Requires curl, tar, and GitHub CLI (gh) for attestation verification. gh does not need a GitHub login for the local bundle check. Unstamped source builds report version: "dev".

Release installer Recommended · macOS / Linux
curl -fsSL https://raw.githubusercontent.com/myrrazor/atlas-tasker/main/scripts/install.sh | sh

The installer only places the binary. If stdout is a terminal and the current directory is already an Atlas workspace, it may offer tracker setup and defaults to No. Set SKIP_INTEGRATIONS=1 to suppress the offer; unattended installs never prompt.

Optional Go install Go 1.26.6+
go install github.com/myrrazor/atlas-tasker/cmd/tracker@latest

Add $(go env GOPATH)/bin to your PATH if needed. Unstamped source builds report version: "dev".

Release files Checksums, attestations, and platform archives
Your first board tracker init then tracker; directory named app yields project APP
mkdir app && cd app
tracker init
tracker
tracker ticket create --project APP --title "Ship first feature" --type task --actor human:owner --reason "first ticket"
tracker board

tracker init then tracker is the workspace path: default project, local checkpoints, Atlas-managed MCP named atlas-tasker for detected agents (no picker), and Home on 127.0.0.1:7432. Restart the coding agent after init so it loads Atlas MCP. Mutation commands resolve identity from --actor, TRACKER_ACTOR, or actor.default; without one, they exit before writing. Include --reason to preserve why the change happened. The getting started guide walks the whole path.

The sensible questions.

What is Atlas Tasker?

Atlas Tasker is a local-first issue tracker and orchestration layer that lives in your code repository. Humans plan on the web board or in the terminal, and AI coding agents work the same tickets through the CLI and MCP, with dependencies, evidence, review gates, and audit history.

What does Atlas Tasker cost?

Nothing. Atlas Tasker is MIT licensed, with no hosted service, account, or paid tier required.

Where is my data stored? Does it phone home?

Tickets are Markdown files in your repository, and changes are recorded in an append-only local event log. A small disposable index keeps queries fast and can be rebuilt at any time. Atlas Tasker has no hosted mode, account system, or telemetry service.

What platforms does Atlas Tasker support?

The release installer supports macOS and Linux on arm64 and amd64. Building from source requires Go 1.26.6 or newer.

How do coding agents use it?

tracker init writes Atlas-managed MCP entries named atlas-tasker pointing at tracker mcp serve --global --tool-profile workflow. Restart the client. A written file is not a live connection. Agents can also drive the CLI with JSON output and stable exit codes. The For agents page covers global serve, pinned --workspace, and the six skill targets.

Do I wire each coding agent by hand?

No. The installer only places the binary; it does not initialize a repo or register agents. After install, tracker init detects Claude Code, Codex, Cursor (including cursor-agent), OpenClaw, and Grok and writes Atlas-managed MCP entries plus the worker skill unless you pass --no-agents. Restart the client. Grok needs its own project-trust prompt before local skills appear. Atlas does not start those agents. Open the repo and ask for the board. There is no second Atlas install inside the client.

How do I get support?

Open a GitHub issue for bugs, setup trouble, documentation problems, and feature requests. Security reports should follow the private reporting instructions in SECURITY.md.

How do I update Atlas Tasker?

Run tracker update --check, then tracker update --yes to verify checksums/attestations and replace the binary. Update replaces the binary only; re-run tracker init in each workspace to refresh managed skills. Use --version to pin a release, or re-run the curl installer.

Why not just use Jira?

Use Atlas Tasker when work should travel with a codebase and coding agents need a tool-native workflow. Jira is still a better fit when your organization needs a hosted, cross-team system with centralized administration.

How do I uninstall it?

Run tracker uninstall to preview Atlas-owned software, then tracker uninstall --yes to apply. Boards, history, backups, and registry pointers stay. Removing the binary by hand also leaves repository data in place.

How do I delete my data?

Delete the Atlas Tasker files from your repository. If Git tracks them, they can remain in repository history until you rewrite that history. The marketing site has no account or hosted copy of your project data to delete.

Useful? Give the repo a star.

Atlas Tasker is free and MIT licensed. A star helps other terminal people find it. Issues and practical pull requests are welcome too.