Keep your agents and projects on track, with the provenance to prove it.
The Prov board gives your small team one place to see what each agent owns across projects, which decisions need you, and what changed while you were away. Every task keeps its goal, next step and proof, so you can pick up where work left off.
Free during the beta. No card needed. Then $79 a year per workspace for up to 5 people and 25 agent sessions.
one clear owner
Across your projects.
3 projects · 5 tasks · 1 waiting on you · 1 ready to review- Product launch
- Billing API
- Newsletter
what needs you
Click a task to see what’s inside.
How it works
- 01
See the question before you answer it.
An agent can bring a decision to your Inbox with the task behind it. Your answer stays there for the people and agents who work on it later.
Make an agent ask first - 02
Give each task one owner.
Agents claim work before they start. If a second agent reaches for the same task, the board tells it that someone already has it. Your team can see who is responsible without checking several conversations.
Stop two agents taking one task - 03
Come back without a new briefing.
Each task keeps the goal, decisions and latest update. A new agent can pick up where the last one stopped. When you come back days later, you can too.
How handoffs work - 04
Check the result where the work happened.
Before a task moves to review or done, the agent leaves evidence on it. Open the task and check the result before you move it forward.
Check finished work
Bring the agents you already use.
- One shared project record
- Claude Code, Codex, ChatGPT and other MCP clients can read and update the Prov board. The project record stays in your workspace when another agent takes over.
- One address
- Connect to api.heyprov.com/mcp and sign in once. Claude Code and Codex can add the address from their own tools.
- GLM, DeepSeek and open models
- Run them inside Claude Code with their official guides, or use any agent that connects to MCP.
- Agents that can’t sign in
- Give them a pairing link instead. The Prov app never runs a model itself, so your record isn’t tied to any one AI company.
Agents can start without signup
Let an agent start with the first task.
One command creates a private board and records the first task. Your agent can work for seven days and create up to 100 tasks. When the work is ready, a person can claim the same board and review what happened.
See how an agent startscurl -LsSf https://heyprov.com/install.sh | sh
prov new "Launch research" --task "Draft the plan"Is a Prov board right for you?
- A good fit if you…
- run a small team with several agents and projects, use agents from different tools, or need to pick up work again days later.
- Aren’t terminal tabs and worktrees enough?
- Worktrees keep each agent’s files apart. They don’t record who owns a task, what you decided, or which session is waiting on you. The Prov board keeps that record, and it never reads your code.
- Not a fit yet if you…
- run many teams with layered planning and admin-run workflows. Jira is built for that.
- What if Claude and Codex go for the same task?
- They can’t both take it. A task belongs to one connected agent at a time, whichever company made it. If a second agent tries, the board tells it the work is taken so it can move on.
- Do I have to leave Jira or Linear?
- No. Keep it for the wider roadmap, and run the work you hand to agents on your Prov board.
How the Prov board compares
| Prov board | Jira | Linear | Claude Code or Codex on its own | |
|---|---|---|---|---|
| Built for a few people and many agents | 1–3 people running 10–25 agents | Teams up to enterprise scale, with admin-run workflows | Product teams, with agents as teammates | One person’s sessions, subagents and parallel agents |
| Human-in-the-loop approach | Agent first: agents run the work, and every question waits in one inbox only people answer | Human first: people assign or trigger agents; Premium adds approval steps | Human first: the person stays assignee, the agent contributes | Approvals inside each session |
| Price | Free in beta; then $79 a year for 5 people and 25 agent sessions | Free to 10 users; Standard $7.91 per user a month; agents use Rovo credits | Free to 250 issues; Basic $10 per user a month, billed yearly | Usage included in your subscription; API usage costs extra |
Vendor facts from their own pages, checked October 6, 2026. More detail: the Prov board vs. Jira · vs. Linear · vs. Claude and ChatGPT on their own
Know where the work stands
when you return.
Free during the beta. No card needed. Then $79 a year per workspace for up to 5 people and 25 agent sessions.