BETAA system of record for AI-assisted workflows

Task tracking built for human + AI collaboration.

Handover keeps tasks honest across runs. Plans, work, and reviews live in one timeline — your agents propose, you approve, and the next agent picks up exactly where the last one left off.

Create a free account→Sign in

Connect from◇ Claude Code◇ Cursor◇ Windsurf◇ ChatGPT

◆ Hosted MCP · 60+ tools

~/.config/claude/mcp.json

# 1. Sign in and grab your config from # https://handover.run/mcp # 2. Paste it into your MCP client # (Claude Code, Cursor, ChatGPT, Windsurf…) { "mcpServers": { "handover": { "url": "https://handover.run/mcp" } } } # 3. Hand off, review, repeat. → agent created task task_a1b2c3 — "Add reporting feature"

What you get

A shared workspace for your agents and your team.

Three primitives. One source of truth. Built for engineers who run multiple agents in parallel and still want to know who did what, when, and whether it's been reviewed.

Claude CodeCursorWindsurfChatGPTMCP

01 · MCP

Works with the editor you already use.

Sign in, copy your config from handover.run/mcp, and paste it into Claude Code, Cursor, ChatGPT, or Windsurf. Your agents speak the Handover protocol from the next prompt onward.

Claudesubmitted plan2h ago

Mariusapproved plan1h ago

Claudehanded offjust now

02 · Audit

Full history and ownership, tracked automatically.

Every plan, handover, approval, and return is logged with an actor and a timestamp. Resume a task three weeks later and the context is still there.

✓Reproduce the bug

✓Patch the callback handler

Add an integration test

Ship behind a flag

03 · Steps

Step timelines for multi-step work.

Long tasks break into ordered steps so agents can check in mid-flight, and you can review a partial result without losing the next-action.

Full MCP parity

Anything you can do in the app, your agents can do over MCP.

Handover ships an MCP tool for virtually every action in the web UI — creating tasks, handing off, reviewing, returning, commenting, managing actors. Drop the endpoint into Cursor, Claude Desktop, or ChatGPT and you can operate the whole product from a chat window.

◐In the appweb UI

＋Create a new task

▤Read & list tasks

→Hand off to another actor

◈Submit work for review

✓Approve or return

≡Add steps to a task

◯Comment on a handover

⊕Manage actors & groups

1 : 1

◆Over MCPtool call

◆handover.create_task✓

◆handover.list_tasks✓

◆handover.handover✓

◆handover.submit_for_review✓

◆handover.approve / return✓

◆handover.add_steps✓

◆handover.add_comment✓

◆handover.upsert_actor✓

60+MCP tools, covering every action in the frontend.

~100%Feature parity between the UI and the MCP surface.

0Frontend visits required to ship — agents can run end-to-end.

Operate fromClaude (Desktop & Code)CursorChatGPTWindsurf

In the app

Every handover reads like a memo, not a chat log.

Three section headers do the work: what was done, what to check, and next steps. Approve or return in one click — the next agent sees your response as the next handover.

◈Triple-redundant status — colour, glyph, label. Survives every screenshare, every reader, every theme.

↩Return with a comment — your prose becomes the next agent's brief.

⌘KKeyboard-driven — review an inbox of approvals without leaving the home row.

handover.run / tasks / task_a1b2c3

Add reporting feature to dashboard

tasks · task_a1b2c3

◈ Needs review● Highupdated 2h ago

Claude◆ AI agent→ Marius

2h ago

What was done

Wired the new reporting card into the dashboard route, ported the date-range picker from settings, and added a CSV export endpoint behind a flag.

What to check

Open /dashboard?range=last_30_days and verify the totals match the SQL in the handover. CSV export is gated by HANDOVER_REPORTING=1.

Review handover from Claude · Submitted 2h ago

Get started

Bring the system of record to your next agent run.

Sign up, copy your config from handover.run/mcp, and your existing Claude Code or Cursor session is already speaking the Handover protocol.

✓ Free during beta — no credit card.
✓ Hosted MCP endpoint. Nothing to install or run.
✓ Works with the agents you already run.

Create your account

Just name, email, and a password to get started.

By signing up you agree to our terms.
You can delete your account at any time.