---
url: /handbook.md
description: >-
  DutyGraph maps how a company actually works, finds where AI can help, and
  governs the authority of the agents built on that work.
---

# DutyGraph

**The aim: every AI agent permission traced to the work a real person actually does.**

**Built today:** DutyGraph keeps each agent within its owner's existing access and the company's rules, routes it to human approvers and records every decision. Tying each permission to the specific work that needs it is in progress.

DutyGraph does two things with one model of the company:

1. **Operating Intelligence: it maps how the company actually works.** Research, the sponsor, the executives and the employees each add what they know. DutyGraph turns it into a Company Repo of Task Cards: the duties, steps, systems, owners and rules behind the work.
2. **Delegated Authority Governance: it governs the agents built on that work.** An agent's proposed permissions are checked against its owner's existing access, the System Catalog and the company's rules. The people who already approve human access decide. Every decision is recorded on a hash-linked ledger. The company's own identity tools provision the agent.

The Task Card joins the two halves. The goal is that the same card that explains the work also justifies each permission an agent holds. Today that link is **Partial**. See [The key join](governance/the-key-join.md).

**Three ways in.** A company can start from its strategy (top-down), from one department (pod-down) or from one agent request (bottom-up). All three fill the same model. Top-down is built today; the other two are Planned. See [Three ways in, one model](start-here/three-ways-in.md).

::: info
**What these docs describe.** Each page labels its subject as **Built** (in the code, with tests), **Partial** (some Built, the rest Planned), **Simulated** (the screen works, but no real outside system is connected) or **Planned** (written down, not built). See [Built and planned](start-here/built-and-planned.md).
:::

## Find your way

| Section | What it covers |
| --- | --- |
| [Start here](start-here/index.md) | What DutyGraph is, the three ways in, the loop it follows, the 9-step journey, who uses it, and what is built |
| [Map the company](discovery/index.md) | Company research, the sponsor packet, the executive kickoff, team asks, the Company Repo, Task Cards and the Work Map |
| [Find opportunities](opportunities/index.md) | Hours and AI modes, possible constraints, Strategy, Route and Reports |
| [Build agents](agents/index.md) | The agent builder, manifests, the plan shared with the person, and exports |
| [Govern agents](governance/index.md) | Holdings, the System Catalog, rule packs, the approval chain, the ledger, the gateway and audit evidence |
| [Run engagements](advisors/index.md) | Accounts, the Inbox, payment, Living DutyGraph, DutyGraph Agent and phones |
| [Connect](connect/index.md) | The MCP servers and integrations |
| [Architecture](architecture/index.md) | How the system fits together, the data model, and where AI is used and where code decides |
| [Trust center](trust/index.md) | Security and data handling, SOC 2 status, subprocessors, data retention, evidence and provenance |
| [Roadmap](roadmap/index.md) | What is planned, and where DutyGraph is going |
| [Reference](reference/index.md) | Glossary and FAQ |

## Who these docs are for

| Reader | Start with |
| --- | --- |
| Advisors and forward-deployed engineers | [Run a company in 9 steps](start-here/run-a-company.md) |
| Company leaders and sponsors | [What is DutyGraph](start-here/what-is-dutygraph.md), [Map the company](discovery/index.md) |
| Identity and access management (IAM) teams | [Three ways in, one model](start-here/three-ways-in.md), [Govern agents](governance/index.md), [Set up governance](governance/set-up-governance.md) |
| Department heads | [Three ways in, one model](start-here/three-ways-in.md) (pod-down discovery is Planned) |
| Internal Controls and audit | [Rule packs and SoD](governance/rule-packs-and-sod.md), [Audit evidence](governance/audit-evidence.md) |
| Heads of AI and platform teams | [Exports and runtimes](agents/exports-and-runtimes.md), [Agent gateway](governance/agent-gateway.md) |
| Employees | [Team asks](discovery/team-asks.md), [Request from your AI](governance/request-from-your-ai.md) |

::: info
**Status.** Top-down discovery, opportunities, the agent builder, the evaluator's ceiling and rule checks, the approval chain, the ledger, the gateway check and the evidence export are **Built**. Limiting an agent's scopes to what its Task Cards need is **Partial**. The identity and business systems in the samples are **Simulated**. Pod-down and bottom-up entry, live identity connections and provisioning are **Planned**. See [Built and planned](start-here/built-and-planned.md).
:::
