Skip to content

Built and planned ​

This page is the status reference for every other page in these docs. When another page and this page disagree, this page wins.

Last checked: October 15, 2026, against the development branch at commit 699f5236.

DutyGraph labels every capability in these docs.

  • Built: in the code, with tests.
  • Partial: some of it is Built and the rest is Planned. The row says which part.
  • Simulated: the screen works, but no real outside system is connected.
  • Planned: written down, not built.

WARNING

"Built" means the code exists and its tests pass. It does not mean a customer runs it in production. These docs make no claim about customers, results or certifications.

Map the company ​

CapabilityStatusPage
Add a company by its website; research with sources; Company SnapshotBuiltCompany research
Research refresh plans, Refresh now, Keep research currentBuiltCompany research
Live news and competitor feedsPlannedCompany research
Sponsor packet: private link, voice with captions, team CSV, Here's what we heardBuiltSponsor packet
Send me a test before the real sendPlannedSponsor packet
Kickoff agenda, invitations, calendar file, pre-read pageBuiltExecutive kickoff
Kickoff transcriptPartial: paste only; file upload, audio and recorder import are PlannedExecutive kickoff
Team asks on a phone, hours as tap ranges, answers stored "as understood"BuiltTeam asks
Company Repo with eight file states, To review, Download your companyBuiltCompany Repo
Rule packs and an import receipt compiled into the RepoPlannedCompany Repo
Task Cards with [H], [A] and [S] steps and AI modesBuiltTask Cards
Work Map: value chain, work flows, Task Cards, people, RACIBuiltWork Map
Top-down discovery designed around a functional organizationBuiltThree ways in, one model
Pod-down discovery: one department sponsors a bounded unitPlannedThree ways in, one model
Bottom-up task capture when an agent request has no adequate Task CardPlanned. The request tools are Built; capture, manager confirmation and application-owner validation are notJust-in-time task capture
Matrix organization shapes; fully AI-run capture sessionsPlannedMap the company
Mission lineage: each task linked to why it existsPlannedTask Cards

Find opportunities ​

CapabilityStatusPage
Reported hours, the code-only AI-fit pass, proposed agentsBuiltOpportunities
Hours with an agentSimulated: authored values in the sample companies onlyOpportunities
Possible constraints, Strategy exhibits (RACI, Theory of Constraints, SWOT / TOWS), DutyGraph AgentBuiltOutcomes and Strategy
Other framework exhibits; scheduled weekly invitationsPlannedOutcomes and Strategy
Route: the cheapest change that can work, by codeBuiltRoute
Reports: Prepare, Saved, Deliver, first view, Print / Save as PDFBuiltReports

Build agents ​

CapabilityStatusPage
Proposed agents, builder, versioned and hashed manifests, freshnessBuiltAgent builder and manifests
Private 30-day agent plan link for the personBuiltShare the agent plan
Five export formats with README and GAPS filesBuiltExports and runtimes
Copilot declarative agent package; hosted agents on the company's vendor accountPlannedExports and runtimes

Govern agents ​

The aim is that every permission an agent holds is necessary for confirmed work, justified, approved and controlled. Today DutyGraph keeps an agent within its owner's holdings and the company's rules, and records each human decision. Proving that each scope is the minimum the work needs is not built yet.

Deciding an agent's authority ​

CapabilityStatusPage
Evaluator: checks each requested scope is well formed and in the System Catalog, removes scopes the owner does not hold (rule CEIL-01), applies the agent rules, gives one state per scopeBuiltThe key join
Task Card requirement check (rule CEIL-02): keep only scopes the selected Task Cards needPartial. The rule is written in the default rule pack. The evaluator does not apply it: it checks only that each Task Card id exists. A requester can ask for any catalogued scope the owner holdsThe key join
What a Task Card is taken to needPartial and coarse. The scope lines listed on the Task Card; if none, every catalog scope of the systems it names. A system with no listed scopes gets generic read and update scopesTask Cards
Step-level business capabilities; entitlement mappings validated by application ownersPlannedTask Cards
SoD check on the person, the agent alone, and the person plus all their agentsBuiltRule packs and SoD
Approval chain: manager, role owner, Internal Controls, IAM; 3 decisions per stepBuiltApproval chain
Hash-linked ledger with Verify chain; optional server sealBuiltLedger
Agent Authority Record: one view of an agent's purpose, owner, Task Cards, scope calculations, decisions and lifecyclePlannedRoadmap
Human-delegated agents: authority within the named owner's holdingsBuiltHoldings
Service-owned or department-owned agents with their own authorization modelNot supported. A future considerationRoadmap

Running and controlling an issued agent ​

CapabilityStatusPage
Gateway check on each routed call: agent status, expiry, issued scope and tool-to-scope match; the credential must match the issued manifestBuilt. Only calls routed through the gateway are checked. The business systems behind it are SimulatedAgent gateway
Gateway activity record for each routed callBuiltAgent gateway
Per-action human approval ("Ask before supervised calls") at the gatewayNot built. The gateway has no approval gate, although the app's runtime table names one. Only a runtime's own approval feature, where one exists, can provide this todayAgent gateway
Revoke an issued agent; end an event-bound agentBuilt, as manual actionsRequest a governed agent
Owner departure suspends the owner's agentsNot built. Marking a person as left updates their status, revokes unused invitations and disconnects their phones. Their issued agents and credentials stay active until someone revokes themRoadmap
Re-check issued agents against the owner's current holdingsPlanned. The gateway checks the scopes issued at approval, not today's holdingsRoadmap
Dependency revalidation when holdings, Task Cards, rules or mappings changePartial. Manifest freshness (Fresh, Out of date, Broken) and credential expiry are Built. Re-assessing issued agents is PlannedAgent builder and manifests

Evidence and the company's identity systems ​

CapabilityStatusPage
Register and evidence export (CSV and JSON) for agents requested through DutyGraphBuiltAudit evidence
Agent inventory reconciliation: match the company's real agent population to governed agentsPlanned. The "found outside DutyGraph" identities in the sample are illustrativeAudit evidence
Full zip evidence packet with a standalone verifierPlannedAudit evidence
Set up governance for any company (6 packs)BuiltSet up governance
Holdings from an uploaded access-export CSVBuiltHoldings
Live Okta, Entra, Saviynt and ServiceNow connectionsPlannedIntegrations
Provisioning the agent identity in the company's IAMPlanned; today a "For IAM to do" card. The company's IAM provisionsRequest a governed agent
Decision Export (the approved decision, sent to IAM or IT service management tools) and Event Import (provisioning, revocation and other events, reconciled with what was approved)PlannedIntegrations

Run engagements ​

CapabilityStatusPage
Accounts, outreach with compliance rules, InboxBuiltAccounts and Inbox
Proposal, Stripe Checkout and invoices, activation by webhookBuilt; off until the server has Stripe keysEngagements and payment
Living DutyGraph: scheduler, digests, renewals, client status pageBuiltLiving DutyGraph
DutyGraph Agent: Ask, Review, Edit, Draft an askBuilt; needs an approved AI connectionDutyGraph Agent
Phone layouts on every page; advisor web pushBuilt; push is off until the server has push keysPhones and notifications

Connect, trust and security ​

CapabilityStatusPage
Advisor MCP server (16 tools) and employee request tools, token sign-inBuiltMCP for advisors, Request from your AI
OAuth sign-in for web and mobile AI connectorsPlannedConnect
Run tools and run tracking from OpenTelemetryPlannedRoadmap
Tenant, company and participant isolation with forced row-level securityBuiltSecurity and data handling
Enterprise SSO and MFAPlannedSecurity and data handling
Security certificationsNone claimedSecurity and data handling

How this page is maintained ​

  • An internal implementation-status register backs this page. Engineers update the register when a capability changes, then update this page to match.
  • Each row is checked against the code on the development branch. A row says Built only when the code and its tests exist.
  • When a change moves a row, update the Last checked line at the top with the date and commit.
  • Other pages link here instead of repeating a status. If you find a page that disagrees with this one, correct that page.

INFO

Status. This page lists every capability with one status. Top-down discovery, opportunities, the agent builder, the evaluator's ceiling and rule checks, SoD checks, the approval chain, the ledger, the gateway check and the evidence export are Built. The Task Card requirement check and dependency revalidation are Partial. The identity and business systems in the samples are Simulated. Pod-down and bottom-up entry, step-level capabilities, owner-departure suspension, per-action approval, reconciliation, Decision Export and Event Import are Planned.

Source: gitbook/product/start-here/built-and-planned.md in the DutyGraph repository.

DutyGraph. Pages label each feature Built, In review or Planned.