Skip to content
Conductor

The console

Built to observe, configure and override the agents

The Next.js console is not another tracker. It opens on what needs a person today, shows the grounding behind every agent action, and lets an admin change course with a stated reason.

adminexecutivedeveloper

Today

Decisions waiting, grounded risks, agents in motion, and the Measure Strip: every project's live status mix between barlines.

  • A first-run checklist walks a new tenant from model configured to first plan approved in eight checks.
  • Developers see their own work; executives see the organisation.
TodayThe score: one measure per project
Hover a measure for its status mix
CompleteIn reviewIn progressBlockedAssignedBacklog> at risk of its target date
admin

Decisions

The human-in-the-loop queue: proposed task lists, assignment proposals and gated emails, grouped by run, with a live badge in the nav.

  • Resolve one or a batch; the paused graph resumes on the last decision.
  • Every resolution lands in the audit log with the reason.
DecisionsHarbor data pipeline, planning run
  1. DoneDeveloper profiles
  2. DoneTask plan
  3. Task list approval
    Waiting for you
  4. Assignment proposals
  5. Assignment approval
  6. GitHub provisioning
Proposed task list
3 tasks from the uploaded FRD
Waiting for you
  • TASK-160Ingest schema for sensor feeds4 d
  • TASK-151Offline sync for field reports5 d
  • TASK-152Photo capture with GPS stamp3 d
adminexecutive

Projects

Workspaces with Overview, Tasks, Team, Risks and Updates tabs, launched from a document upload and replanned through change orders.

  • The task inspector shows the grounding behind every completion: the merged PR, the review, the blocked reasons.
  • Force-complete and reopen ask for a reason and refetch, never guess.
Task inspectorTASK-142 Invoice PDF export
#87feature/TASK-142-invoice-pdf-exportOpen
+412 −38 in 9 files
Code review reviewing
reading the diff, the acceptance criteria and the python-fastapi playbook
The completion rule
  1. A linked pull request is merged into the repository's default branch
  2. The latest review covers that exact merged commit
  3. The review did not request changes and found nothing high or critical
  4. The acceptance criteria are marked met
Evaluating…
adminexecutivedeveloper

Workboard

Every task across projects as a board, a grouped list, a sortable table or a due-date timeline.

  • Drag actions exist only where an audited endpoint exists; status still follows the pull request and its review.
  • Deep links open the inspector for any task.
Workboard8 tasks across 3 projects
Backlog2
TASK-153Supervisor sign-off screen—
TASK-160Ingest schema for sensor feeds—
Assigned1
TASK-145Tax rate table adminBO
In progress2
TASK-144Customer statement emailsTR
TASK-152Photo capture with GPS stampAS
In review1
TASK-143Webhook retry queuePN
Complete1
TASK-142Invoice PDF exportPN
Blocked1
TASK-151Offline sync for field reportsAS
adminexecutive

People

The roster, résumé-built skill profiles, and the assignment pools the planner draws from.

  • Import users from Microsoft Entra ID or register them by hand.
  • Profiles assist assignment; they are never a surveillance surface.
PeopleSkill profiles built from résumés
  • PNPriya NatarajanSeniorPythonFastAPIPostgreSQL2 of 3 active
  • TRTomas RuizMidTypeScriptNext.jsTailwind3 of 3 active
  • ASAiko SatoMidReact NativeGraphQL1 of 3 active
  • BOBen OkaforJuniorPythonSQL1 of 3 active
admin

Playbooks

Per-stack coding standards from upload, through the Recon agent's distillation, to activation and enforcement in code review.

  • Activate to embed; deactivate to purge. The review only ever sees active playbooks.
  • Re-distill after editing the source document.
PlaybooksPer-stack standards
  • python-fastapiActive14 rules
  • nextjs-reactDistilled, pending activation11 rules

Activate to embed; deactivate to purge. Code review sees only active playbooks.

adminexecutive

Mail

Every agent email thread with reply-gap triage: what went out, what came back, and what is waiting on a person.

  • Inbound replies are threaded and embedded into memory.
  • Feedback and alerts can be gated; the digest auto-sends.
MailAgent threads
  • FeedbackTASK-143 Webhook retry queue: review feedbackReplied 2 h ago
  • AlertMeridian field app: risk alertWaiting 3 d
  • DigestWeekly digest: Atlas, Meridian, HarborSent
admin

Activity

The full audit log: 77 event types covering every autonomous action and every manual override, each with actor, timestamp and reasoning.

  • Filter by actor, action or target.
  • Overrides are red-pencil entries; agent actions are blue.
ActivityThe audit log
  • Agenttask.auto_completedTASK-142

    PR #87 merged into main; review 9f3c2a1 approved; acceptance met.

  • Userassignment.approvedTASK-152

    Aiko owns the mobile capture flow.

  • Agentrisk.detectedMeridian field app

    TASK-151 blocked 6 days; milestone in 5 days.

  • Overridetask.force_completedTASK-138

    Delivered via hotfix branch outside the PR flow.

admin

AI usage

Token usage per agent, flow and project, captured by one LangChain callback and never a phantom row.

  • Know what the planner costs versus the reviewer.
  • No prompt or response text is stored.
AI usageTokens by agent, last 30 days
  • Planner184,300
  • Code review96,120
  • Communication41,800
  • Profile22,450
  • Recon17,900
  • Assignment12,600
  • Change planner9,300
admin

Settings

Change the model provider, GitHub, SMTP or IMAP configuration at runtime with inline verification. Secrets are encrypted at rest and never serialised back out.

  • LM Studio locally or OpenAI, by one setting.
  • Per-group saves; a bad value is refused before it is stored.
SettingsModel
Provider
LM Studio
Base URL
http://host.docker.internal:1234/v1
Chat model
qwen2.5-7b-instruct
VerifyReachable, 1 model loaded

Three roles, enforced on the backend

Admin
Everything: decisions, projects, people, playbooks, mail, activity, AI usage, settings.
Executive
Today, projects, workboard, people and mail. Reads, never resolves.
Developer
Today and My work: only their own tasks, enforced on the backend.

See Conductor on your own repositories

Book a walkthrough of a live tenant, or request a pilot on your GitHub organisation.