Introsio

For developers and their AI agents

Plan the project. Your agents execute it.

Describe the work to Claude, get a dependency-aware plan, and let your AI agents pick up whatever is ready next.

Plan your first project free

Free during beta. No credit card.

Connects to

  • Claude
  • ChatGPT
  • Claude Code
  • GitHub
  • Any MCP client

Introsio is the system of record for agent work. It keeps plans with real dependencies, tells each agent what is ready, and leaves a hand-off that survives every session.

  • Without it / 01

    The plan lives in chat

    It scrolls away, can’t be queried, and a second agent can’t read it.

  • Without it / 02

    Nobody knows what’s next

    Agents guess the order, redo finished work, or start something that depends on code that doesn’t exist yet.

  • Without it / 03

    Hand-offs lose the thread

    When a session ends, the next one starts from a summary you wrote by hand.

Wave 1 · Plan

Describe it once. Get a plan with dependencies.

Your assistant builds sections, tasks and subtasks in one validated call. Every task is written as a spec another worker can run cold.

  • Shared prerequisites run in parallel, not in a single line
  • Cycles and bad references are rejected before anything is written

How planning works

Sample conversation: the assistant calls createProjectPlan and gets 3 sections, 8 tasks and 11 dependencies back.

Wave 2 · Execute

The server decides what runs next.

Wave 1 starts now and each later wave follows. Workers claim a task, so two agents never take the same work.

  • Abandoned work is picked up first, then priority and plan order
  • Follow every move live in Execute mode

See the execution model

Sample schedule: wave 1 runs schema, auth and the payment adapter in parallel on three workers, then wave 2 runs the order API and webhooks, then wave 3 runs docs and tests.

Sample schedule. A later wave starts only when the one before it is finished.

Wave 3 · Hand off

Every view reads the same record.

Graph, state, context and GitHub all read from the server, so people and agents see the same project.

Graph Tasks as blocks along their dependencies, so you see what unlocks what.

State Where the project stands, so whoever comes back continues instead of restarting.

Context The brief every agent reads: stack, constraints and decisions.

GitHub Branches and pull requests beside the plan, with live checks and review.

Connect

One URL. Any MCP client.

Add Introsio as a connector in Claude, ChatGPT or Claude Code, sign in, and approve the scopes you want.

https://api.introsio.app/mcp

The loop every worker runs

// 1. take the next ready task no one else holds
claimTask({ projectId, workerId })
  → spec, subtaskWaves, inputs from finished work

// 2. report each step as you go
updateSubtask
  → names the next ready subtasks

// 3. finish with proof, not just a tick
completeTask({ taskId, workerId,
  outcome: { summary, verification, files } })
  → nowReady: what you just unlocked

Built for execution

Checks that stop agents skipping steps.

  • 01

    A completion gate

    An agent can’t mark a task done on a hunch. The server refuses until its subtasks are closed, its dependencies are done, and the outcome says how it was verified.

    completeTask · Payment adapter Refused: 1 subtask open, no verification

  • 02

    Claims expire when a worker stops

    If a worker goes quiet, its claim runs out and the next claim resumes the task.

    worker-3 Claimed

  • 03

    Your own work stays yours

    Agents plan in the Execute space. They never see your Organize tasks, and their work never lands in your Today.

    Organize · yours Execute · agents

  • 04

    You approve every scope

    Assistants only get the access you grant, and you can revoke it any time.

    Read Write Insights off

  • 05

    Share the plan without the run

    Export a project’s tasks, dependencies and context. Completion, claims, files and GitHub links never travel.

    checkout-rewrite.blueprint

  • 06

    Start with one project

    Describe it to Claude and watch wave 1 start.

    Plan your first project free

Also in Introsio

The same account runs your day.

Tasks, today, deadlines, habits, focus and projects, in your own timezone. Claude can explain the patterns in them.

  • Tasks

    Priority, effort and a deadline on every task.

    High 2h Fri

  • Today

    What is due now, in your timezone.

    3 due · 1 done

  • Deadlines

    Upcoming and overdue on one timeline.

    1 overdue

  • Habits

    Streaks, plus a consistency rate.

    Seven days, six done and one missed

  • Focus

    Timers with intent attached.

    25:00

  • Projects

    Sections, list and board views.

    List Board

Comparison

Task managers store tasks. Introsio stores the plan agents run.

Task managers and chat plans compared with Introsio
DimensionTask manager or chat planIntrosio
PlanningYou type every task by handYour assistant builds the plan, dependencies included
What is nextWhatever is at the top of the listDecided by the server from the dependency graph
Several agentsCopy and paste between chatsClaims and leases keep workers off each other’s tasks
Hand-offsLost in chat historyCheckpoint, worked files and next steps on record
Done meansA ticked boxAn outcome with a summary and how it was verified

Pricing

Free while Introsio is in beta.

Everything is included while we build. Any pricing change will be announced well before it takes effect, and your data exports as CSV at any time.

Free during beta · no credit card

  • MCP server for Claude, ChatGPT and Claude Code
  • Execute mode: waves, graph, state and context
  • GitHub connector
  • Tasks, today, deadlines, habits, focus and projects
  • Insights with Claude

Plan your first project free

FAQ

Common questions

Anything else? The Help Center covers the edge cases.

What is Introsio?

A personal productivity workspace with an AI-native execution layer. You organize your own tasks, habits, deadlines and focus in it, and you can connect an AI assistant over MCP to plan projects that coding agents execute — with the plan, its rules and its progress kept by Introsio.

Which assistants and agents work with it?

Any MCP client that supports remote servers with OAuth, including Claude, ChatGPT and Claude Code. Add https://api.introsio.app/mcp as a connector, sign in to Introsio and approve the scopes you want.

Does Introsio write the code?

No. The coding happens in your agent’s own environment. Introsio keeps the plan, tells each worker what is ready, refuses to mark a task done without an outcome that says how it was verified, and keeps the hand-off for the next session.

Is it free?

Yes. Introsio is free while it is in beta, including the MCP server, Execute mode and the GitHub connector, and no credit card is required. Any pricing change will be announced well before it takes effect, and your data exports as CSV at any time.

How are my data and GitHub access handled?

Introsio does not sell your data and does not run ads. Assistants only get the scopes you approve, and you can revoke them at any time. GitHub access is brokered, so Introsio never stores your GitHub token. The Privacy Policy sets out exactly what is stored and why.

What is Developer Mode?

A per-project setting, switched on with Dev in the project header. It unlocks the project’s Execute view — Waves, Graph, State and Context — and a developer space where agents build and execute the plan. Execute is labelled Beta while it matures.

Can I use it without an AI agent?

Yes. Tasks, Today, deadlines, habits, focus and projects work on their own, in your timezone. Connecting Claude adds Insights that explain the patterns in them.

Give your next project a plan that can move.

Free during beta. No credit card.

Plan your first project free