Loading…

Delivery Loop

Sign in to your delivery workspace

Guided setup

New project

Delivery Loop control panel

Overview
Stories
Agent
Feedback
Connect
Settings

Admin console

Platform operations across every organization — analytics, users, structure, and support. Distinct from the user app, which is about a single customer's delivery.

Usage & metrics
—
Daily report

Per-day activity across the whole platform. DAU = unique users active that day.

By organization

Messages, message length, AI replies, stories, and completed work per org (30 days).

Most active users
Sync logs

Every canon→Azure DevOps sync run across all orgs — newest first.

Users & access
Plans & billing
—
Agent runtime
—

Runs agent sessions on your Claude subscription (flat-rate). Stored as CLAUDE_CODE_OAUTH_TOKEN. Once the token + Cloud Run job are set, ▶ Work this runs entirely in-site.

Structure & defaults
Support
Per-user activity & access — help customers get unblocked.

Share an idea

Describe what you'd like or what's not working. I'll ask a couple of quick questions, then log it in Azure as a scored, time-estimated story — recorded for review. No technical detail needed.

Shareable feedback link

Give non-technical people a link to chat their ideas straight into this project — no account needed.

—

Agent

Claude Code working on this project — see if it's active, and guide it with replies.

How your agent works & how to steer it
What it does

Given only your token, a coding agent pulls Ready for dev stories through our API, builds them, moves each In progress → Complete (auto-timed), and posts questions and notable decisions back on the story. It all streams into the live activity feed on Setup.

How you steer it
  • Standing guidelines (build cadence, conventions, deploy/test) → Setup ▸ AI ledger — fed to the agent via config.json.
  • Answer its questions → Stories (the Pending queue) or Needs you.
  • Nudge a live session — reply in its terminal below (e.g. “push”, “continue to the next Ready story”).
— asks the connected machine to build & report an update summary.
Ready to work — fetch a story into the terminal
Select a project to see its Ready queue.
Session history

Users

Invite people into your account by email and share an org — or just specific projects — with them. A project invite adds them to that org with access to those projects only.

Loading…

Organizations

An org holds the master credentials (Azure DevOps PAT + OpenAI key) its projects share.

Add an organization

You only need an Azure DevOps PAT (Work Items: read/write) to start — the AI runs on our platform LLM out of the box. Add your own OpenAI/Anthropic key any time to use your preferred model and lift the free monthly limit. Storage is only needed for feedback sources.

Credentials
Bring your own key for unlimited AI on your chosen model. Leave blank to use ours — capped monthly on the free plan.
Lets the loop push design assets and “wire-up” PRs into your product repos. Skip if you’re not connecting GitHub.
A project can override this with its own Teams app under its Connect → Teams; otherwise it uses the org's.
Run different models across different projects from one hub — set the default here, override per project.

Dashboard

Delivery over time — hours, throughput, estimates vs actuals.

Loading…

Delivery overview

—
▶ Run loop ▾

Pulls Teams/email feedback into scored stories on every active project.

Needs your input

Work through the open questions across your projects — answer, assign, or skip.

Checking…

Agent feedback loop

Issues, gotchas, fixes and tips shared between the agents working your projects — so a problem one agent hits helps the next one avoid it. Platform items reach every agent; project items sync to Azure DevOps.

Loading…

New project

Stories

⋯ Actions

Every story in this project. The New queue is what needs your review — set a priority (P1–P4) on each card, or open one to answer its questions and move it to Ready for dev.

Loading…

Run the loop — no project selected

Each action runs against the selected project with its org's keys, and shows live output below.

output appears here…

My profile

Your personal settings — separate from an org's shared keys. These identify you when the tool acts on your behalf.

My Azure DevOps PAT — so your work is credited to you

Add your own PAT per org. Then anything you do (answers, comments, claiming a story) is attributed to you in Azure DevOps and can be assigned to you — instead of the shared org identity. Required for members to act on a shared org's project. how to get one ↗

—
My GitHub access

A personal GitHub token (scope: repo) tied to you — used when the tool opens PRs or connects repos on your behalf. how to get one ↗

Connecting a repo to an agent and publishing its docs live under a project — open a project and use Connect in its left menu.

Unsaved changes — save this project's connections

Connect

All your 3rd-party connections for the feedback loop — Azure DevOps today; Teams, email and other feedback sources plug in here too.

Feedback connections

Feedback connections

Where feedback comes in. A source you haven't set shows its setup here (amber ○); once configured it collapses to ✓ (green) — click a row to re-open and edit. Only connect what you use.

Azure DevOps — optional
✓ via org keys
○ GitHub — wire-up PRs
▾

The GitHub token lives on the org. This sets the repo for asset/wire-up PRs.

○ Docs store — blob container (needed for Teams/email/docs)
▾
○ Teams channel · setup & access ↗
▾
Reads channel messages read-only (app-only, no sign-in) and turns them into stories. The Teams app (tenant + client + secret) lives on the org — set once, used by every project. Here you just pick this project's channels.
—
The client secret for an override goes in this project's Connect → Keys as a TEAMS_CLIENT_SECRET override.
In Teams: ⋯ next to a team → Get link to team (adds it and lists its channels), or ⋯ next to a channel → Get link to channel. Add links from several teams to pull from all of them. Feedback is tagged by channel name.
Map each channel to what it holds: Chat only (feedback messages → stories), Docs only (documents in the Files tab → context), or Chat + Docs — so a docs channel and a chat channel are pulled correctly.
We resolve the link and pull every file in that folder (and subfolders) into this project's resources. Needs the Teams app to have Sites.Read.All + Files.Read.All (admin consent).
Reading is app-only, but posting must be delegated — authorize once and Delivery Loop posts questions (on Pending user feedback) as you, as a thread, and folds replies back as the answer.
Alternative: a channel Incoming Webhook (if you can't authorize)
○ Support email
▾

Reads a shared support inbox (Graph Mail.Read, reuses the Teams app) → stories. Set up email access ↗

Senders reach this project by putting its subject term above in the email — that's the project link. Unless public, only project members' emails become stories.

When on, a story moving to Pending user feedback emails the assignee (else the owner) the story + latest comments with Reply-To set to this feedback inbox — they can just reply to answer and it threads back onto the story (the [DL#id] subject tag routes it). Needs a feedback address/token set.

○ Notify
▾
Project docs & wiki

Publish this project's summary, the board API reference, and the agent onboarding playbook into your Azure DevOps wiki — so the team and connected agents find them right next to the work items. Delivery Loop keeps its own copy too, and you can re-publish any time to refresh them.


      

      
Keys — inherit from org or override
        
—

        

          
          

        

      

      
    

    
    

      
Agent 

      
Connect a coding agent to build this project from the board — generate a prompt + token, hand it to any agent, and give it standing instructions.

      

      
Connect an agent — quick setup (recommended)
        
One click generates a bounded connect prompt + a scoped .env token. Paste them to your coding agent (Claude Code, Cursor, etc.) and it configures everything and starts working. The token goes in .env, never in the chat.

        

        
 Turbo is a live toggle on the connection status above — flip it any time to fan out sub-agents, or exit to cut it back down.

        

        
Prereq: the org's Azure DevOps PAT must be set — Orgs → this project's org — generated in the SAME Azure DevOps org with Work Items + Wiki (Read & Write) scope.

      

      
AI special instructions
        
Standing instructions fed to the coding agent (via the repo config). Build cadence, conventions, how to deploy/test — e.g. "deploy to staging after each Complete story; run the test suite before pushing".

        
        

        
The AI → you · Decisions

        
—

    



    
    

      
Settings 

      
How this project is estimated, costed, and structured. Changes save with the project.

      
Estimation & cost
        

          

          

        

      

      
Structure & style (bootstrap defaults — only for creating new ADO structure)
        

          

          

          

          

          

        

        
Leave blank for the standard defaults. These drive bootstrap so each org/project can have its own structure.

      

      
Status mapping — Delivery Loop ↔ Azure DevOps
        
Delivery Loop keeps one status vocabulary; this is how it maps to your Azure DevOps columns — round-tripped from what's actually set in ADO. We keep our process and map to yours; we don't change your columns unless you choose to provision below.

        

          
Saved with settings. Blank = auto-detect from your process.

          
Trashed stories are auto-deleted (from here and Azure DevOps) after this many days. 0 = keep forever.

          
When on, feedback/backlog stories are held in Delivery Loop (out of Azure DevOps and the agent's queue) until you Approve them.

        

        
Open a project to see its mapping.

      

      
      
Project details

        
Basics
        

          

          

          
          

          

          

          

          
Set it and the dashboard's Executive Summary opens with your exact words; leave blank and the AI infers one from the roadmap.

        

        

        
Estimation, cost, structure & style are on the Settings tab (left menu).

      

      

    


    
    

      
Connect

      
Plug in a key to unlock each connection. You only need Azure DevOps and OpenAI to start — add the rest whenever you want what they unlock.


      
AI Guidelines — recommended setup & cadence
        
Seeded with our default. Edit it to fit how your team works — connected agents follow these guidelines every cycle for getting feedback in and delivering updates automatically.

        
        

      


      
Keys by organization
        
Keys live on the org — set each once and every project in it inherits them. Pick an org, fill in what it uses (Azure DevOps PAT, the Teams app, GitHub, your LLM key, storage), and save.

        

          
          
          
        

        
Pick an org above.

      

      
Need a project to use a different key (e.g. a client whose Teams or Azure DevOps is outside your org)? Open that project → Connect → Keys and override just that one — everything else keeps inheriting the org.


      

      

      
GitHub accounts
        

        
Add one with gh auth login in a terminal, or save a GITHUB_TOKEN above (that's what powers 🔗 Auto-connect).

      

      
Plan & usage
—

      
    


    
    

  


  
  














  

    

      
Answer

      

    

    

    

      
      
      

        
        

      

        

          
          
          
          

        

      

    

  







  

    

      Add an item to 

    

      
Drop in an idea, notes, or a document — we turn it into scored, estimated stories in this project's backlog. No connected source needed.

      
      
      

        
        
      

      
Text documents (.txt, .md, .csv, .json, .log) are read here and added above. For richer/binary docs or ongoing intake, connect a feedback source (Teams, email, store).

    

    

      

  







  

    
Add a story to 

    

      
Straight onto the board — no AI. Use this when you already know the work.

      
      
      

        
        
        
      

      
    

    

  






  

    
Invite a user

    

      
      
      
      
      
      
Pick an org…

      

    

    

  






  

    
Your private key

    

    

  






  

    
Deep research → stories

    

  





  

    
Plan with AI

    

    

  





  

    
Reconcile with Azure DevOps

    

    

  





  

    
Align states with Azure DevOps