Taskwright

Put your AI subscriptions to work in the background.

Taskwright orchestrates the Claude Code, Codex, Gemini, and Cursor CLIs you already pay for — general-purpose AI agents driving research, writing, analysis, planning, and coding workflows on your own machine, on a schedule, with policy guardrails and silent failover when one provider rate-limits.

BYO-subscription. We never pay for inference — your Claude Max, ChatGPT Plus, Gemini Advanced, and Cursor accounts stay yours, and so do their token budgets.

How it works

Three steps. The agents already live on your machine — Taskwright is the layer that drives them.

1

Install the daemon

A single npm install. The daemon runs locally — no cloud passthrough of your code or your tokens.

$ npm i -g @taskwright/cli
$ taskwright login
$ taskwright start
2

Define an outcome or pipeline

In the web UI: 'Summarize this week's research every Monday' or 'Draft → critique → polish a blog post' or 'Bump dependencies weekly'. Pick which agent CLI to use, or let Taskwright route automatically.

3

It runs in the background

The daemon spawns the agent on your workstation, streams stdout to your dashboard in real time, and escalates to you only when ambiguity, risk, or rate limits demand a human.

Durable runs

Processes die; Taskwright resumes. You don't rerun.

Scheduleable

Cron, interval, webhook. Every Monday at 9am, your agent is already working.

Cross-provider

Route each step to the agent whose subscription has headroom. Silent failover when one rate-limits.

Policy-governed

Deny-lists, escalation patterns, approval gates. Trust is tunable, not binary.

Pricing

Flat monthly fee for orchestration. Your AI subscriptions are paid directly to the providers.

Starter

$20/mo

Solo users setting up their first delegated workflow.

  • 1 daemon, 1 workstation
  • 5 active outcomes
  • All 4 agent adapters (Claude / Codex / Gemini / Cursor)
  • Cron + interval scheduling
  • Real-time log streaming
  • Email escalations

Pro

Most popular
$50/mo

Active solo or two-person teams routing across multiple subscriptions.

  • 3 daemons, unlimited workstations
  • Unlimited outcomes
  • Webhook triggers
  • Cross-provider failover
  • Approval gates + audit log
  • Priority email + chat support

Team

$150/mo

Small teams sharing daemons and policy across knowledge work.

  • 10 daemons, unlimited workstations
  • Shared outcomes + skills library
  • Org-wide policy enforcement
  • SSO (Google, Microsoft)
  • Priority support
  • Onboarding session

14-day free trial on Starter and Pro. No credit card required for the trial. Pricing in USD, billed monthly. Annual plans save 20%.

FAQ

Do I need to bring my own AI subscriptions?
Yes — that's the whole point. Taskwright spawns the Claude Code, Codex CLI, Gemini CLI, and Cursor Agent processes that already live on your machine. You log into each provider once with your existing account; we never see your prompts or your tokens.
Why pay for orchestration if I already pay the providers?
Because the providers don't schedule, retry, persist, fail-over, or escalate. They give you a chat box. Taskwright gives you durable runs, cron triggers, cross-provider routing, and policy guardrails that turn one-off prompts into background workflows.
Where does the work happen?
On your workstation. The daemon spawns the agent CLI as a child process with the working directory you choose — whether that's a code repo, a research folder, a writing project, or anywhere else. Your files never leave your machine; only the run metadata (timestamps, exit codes, stdout for the dashboard) is synced to the cloud.
What if I hit a provider's rate limit?
You configure failover rules per outcome. When Claude Max returns 429, Taskwright can automatically retry with Codex, then Gemini, then Cursor — using subscriptions you already pay for instead of pay-per-token API keys.
Can I run this on Windows / macOS / Linux?
All three. The daemon is a Node ≥22 CLI; the optional desktop shell is Electron with Windows NSIS, macOS DMG, and Linux AppImage builds.