v0.1.0 · public API · open-core

The execution engine
for AI agents.

Track work items, dispatch tasks to agents, log every run, and bill usage — all behind one typed API. The CLI is open-source. The cloud is hosted.

npm install -g backlog Read the docs →

~/my-team — backlog

$ backlog login
  email: alice@acme.com
  ✓ logged in (token expires in 30 days)

$ backlog ws create "acme platform"
  ✓ workspace #14 created · slug: acme-platform

$ backlog item add "Refactor billing" --priority P1
  ✓ WI-42 ready

$ backlog task dispatch WI-42 --agent claude --repo acme/api
  → run #128 queued
  → claude picked up the task...
  ✓ run #128 awaiting_review (1m42s, 12.3k tokens)

$ backlog usage
  plan:      pro
  ai tokens: 12 340 / 5 000 000
  tasks:     1 / 10 000 this month

How it works

A small, opinionated pipeline. Same flow whether you drive it from the CLI, the SDK, or raw HTTP.

Capture

Sync work items from GitHub, Linear, or push them via the API. Each item has a status, priority, and arbitrary payload.

Dispatch

Break items into tasks scoped to a repo. Assign an agent (Claude, Codex, custom) and optionally a sub-scope.

Run

Agents call the AI proxy through your workspace. Every invocation gets a Run with status, summary, artifacts.

Meter

Tokens count against your plan. Stripe handles payment. /usage tells you exactly where you stand.

What you get

A small, opinionated surface for running agents in production.

◆

Work items & tasks

Model what needs to happen, then break it into agent-runnable tasks. GitHub Issues sync out of the box.

▶

Run logs & artifacts

Every agent invocation gets a Run. Status, summary, artifacts, audit trail — queryable, filterable, billable.

AI proxy & quotas

Hit Anthropic through one workspace endpoint. Tokens are metered, capped, and billed per plan.

Built for agent platforms

If you're shipping something agentic, you've already written half of this. We did the other half.

Coding agents

Claude Code, Codex, Cursor agents. Track every PR they ship, the runs that produced it, and what it cost.

Internal copilots

Ops teams running Claude over Slack/Linear. Per-team quotas, audit trail, no key sprawl.

Self-hosted shops

Use the open-source CLI + your own backend. Same OpenAPI contract, your data, your cluster.

Pricing

Free to start. Pay when you actually run things. See the full breakdown →

Free

€0 / month

50 work items
100 tasks / month
3 members
100k AI tokens / month

Get started

Pro

€19 / month

5 000 work items
10 000 tasks / month
25 members
5M AI tokens / month
Priority support

Upgrade

Enterprise

Custom

Unlimited everything
SSO & SCIM
SLA & dedicated support
Self-hosted option

Open-core, by design

The CLI and SDK are Apache-2.0 on GitHub. Point them at our cloud, or run your own backend with the same OpenAPI contract.