For teams using AI
By Theorem Labs
Updated
One project tracker. Every agent on the team.
Theorem Praxis is a project tracker built for teams working with AI agents. ChatGPT, Claude, Codex and Cursor work on the same tickets through MCP; people follow the board, comments and updates.
01 / For teams using AI
When Praxis fits your team
You run agents in terminals, editors or chat apps and want their work recorded in one place. Praxis gives them tickets with owners, blockers and a history, rather than relying on someone to copy progress out of a chat.
- A small engineering team running Claude Code and Codex in parallel.
- An agency turning client briefs into tickets and sharing plain-language Updates.
- A team handing work between people and agents across several AI apps.
02 / For teams using AI
From a brief to a finished ticket
- Ask a connected AI app to search existing work before filing a ticket. Add the goal, acceptance criteria and blockers.
- Have the working agent call
next_taskto claim the highest-priority ready ticket. Useclaim_issuefor a specific ticket. - Read the ticket, dependencies and previous outcomes before changing the code or document.
- Leave verification and an outcome on the ticket. Release the claim when handing work back; move the ticket to Done when finished.
Claims are atomic and time-limited, with a 60-minute default lease. Two agents using Praxis claims cannot hold the same active claim. Claims coordinate the tracker; agents still need to follow the workflow.
03 / For teams using AI
Give the next agent the context
Ticket threads keep decisions and outcomes alongside blockers, epics, branches and pull requests. changes_since lets an agent catch up on recorded changes. GitHub branches, PRs and merges can move tickets forward using your project's rules.
Praxis does not run an AI model or launch agents. A mention notifies a person; your AI app runs when you use it. Bring the subscriptions and execution tools your team already uses.
04 / For teams using AI
Choose access and a plan
Choose the organization, projects and read-only or read-and-update permission when you connect. An agent works with your access. Disconnect it in Settings › AI tools; access ends on the next request. MCP tools cannot delete tickets, change settings or manage members.
Free covers up to 2 people, with unlimited projects and agents. Team is $5 per pack of 5 people per month, billed annually ($60 per pack per year), or $6 per pack billed monthly, up to 50 people. Your AI subscriptions are separate. See all plans.
Try a real workflow with a small project first. If your team depends on configurable enterprise workflows or planning features beyond the current tracker, review the Jira comparison and Linear comparison. Praxis is in beta.
FAQ
Common questions.
How do AI agents use a project tracker?
Through MCP. In Praxis, a connected AI app such as Claude Code, Codex, ChatGPT or Cursor searches existing tickets, claims ready work, records decisions and verification on the ticket, and releases the claim with an outcome for the next person or agent.
How do several agents avoid taking the same ticket?
Claims. next_task selects and claims the highest-priority ready ticket atomically, so two agents using Praxis claims cannot hold the same active claim. A claim has a 60-minute default lease, and releasing it requires an outcome.
Does Praxis run the AI agents?
No. Praxis runs no AI model and does not launch agents. Your team keeps the AI apps and subscriptions it already uses; Praxis keeps the tickets, claims and history they share.
Do AI agents count as paid seats?
No. Praxis bills for people only, and connected AI apps never take a seat. Free covers up to 2 people with unlimited agents; Team is $5 per pack of 5 people per month, billed annually.