Platform capability

ARGENTAGENT WORK TRACKING

Book a demo

Master Control System · by Nyedis

01 / AGENT WORK TRACKING

AI work with no story behind it

Coding agents now work for hours on real stories from your tracker. Their AI spend arrives as one number on a month-end invoice, and the tracker says In Progress and nothing else.

  • 84% of developers surveyed are using or planning to use AI tools in their development process, up from 76% a year earlier.¹
  • 98% of State of FinOps 2026 respondents now manage AI spend, and the report names allocating AI costs to business units as harder than traditional infrastructure.²
  • Gartner predicts over 40% of agentic AI projects will be canceled by the end of 2027, due to escalating costs, unclear business value or inadequate risk controls.³

If you cannot say what a story's AI work cost, you cannot say whether it was worth it.

02 / AGENT WORK TRACKING

Three questions month end cannot answer

  1. Which work did that money pay for? The agents did their jobs, but their spend belongs to no story, no epic and no project.
  2. Who said yes when it ran long? An agent can keep going past what the work was worth, and nobody is asked.
  3. Who allowed that write to our tracker? A comment that just says "Done." from an account named agent tells a reviewer nothing.

Security cannot show who owned the work, and finance cannot tie the bill to a project. See budget and token spend and audit and accountability.

03 / AGENT WORK TRACKING

The one idea: every session declares its story

Argent Strategic Intelligence sits between the agent and the AI model it calls, like Claude or ChatGPT. When the session says which story it is on, every priced request it sends is booked to that story.

  • Developers run their coding agents in ADE, the Argent Development Environment: an agentic harness for agent developers and engineers, built to work hand in hand with Argent. ADE sends the session's model traffic through the Argent gateway.
  • Stories come from the trackers your teams already use. Argent connects to Jira, GitHub, GitLab, Linear and Notion, and a person sees only their own department's stories.
  • Each request is priced on Argent's signed ledger, and its tokens and cost are booked to the story, then rolled up through its epic to its project. A whole project's AI spend is one number. See Identity Tethering for how each agent is tied to a named person.
04 / AGENT WORK TRACKING

How it works, step by step

  1. Pick. At launch in ADE, the developer picks the story from the active sprint, recent stories or a search by title. They can also create a new one; if a similar story already exists, ADE offers to join it.
  2. Declare. The agent tells Argent which story it is on with the tool argent_story_declare. A sub-agent can declare its own, and its spend stays separate. Argent accepts only a story the person can see.
  3. Book. Argent books each priced request of the session to that story, once. ADE shows the developer what the story has cost so far.
  4. Warn. In the Argent console, an admin sets a cap on the story, its epic or its project. At 80 percent, ADE warns the developer.
  5. Hold. At the cap, with Pause chosen, the turn in progress finishes and the next prompt of every session on that story is held. The Argent gateway also checks the story's budget as a second line.
  6. Lift. The budget owner decides in Argent: Raise cap, or One-time overage. The hold ends and the prompt goes through.
  7. Write back. When the agent commits, ADE writes one comment: what changed, which files and what this stretch cost. By default the developer approves it; then, once an admin has switched write-back on, Argent removes secrets, signs it and posts it to the story.
  8. Close. When the work is done, the developer closes the story in ADE. Argent moves the card to Done in the tracker, once an admin has switched that on, and signs that too.
05 / AGENT WORK TRACKING

What the operator sees

WhereWhat it shows
ADE launch pickerA story row with active sprint, recent, search by title, and skip (unassigned).
ADE board cardThe story's Argent cost, the share of its budget spent, and a held chip at the cap.
ADE notices"reached 80 percent of its budget", then a banner: "held: <story> is at its budget. ask the budget owner for more."
Comments waiting for youEach write-back comment, with approve and post, save edit, or discard.
Argent console: WorkWork items, project groups and trackers. A story's Budget section shows a meter (Within cap, Near limit, Over cap, Paused at cap), Set cap, Alert at 80%, At 100% Pause or Alert only, and Lift.
Argent console: Tracker syncA signed trail row per write: person, tracker item, COMMENT, STATUS CHANGE or CLOSE, and POSTED, RETRYING or REFUSED BY TRACKER.
Your trackerThe comment from the organization's Argent bot, with the footer "Posted by Argent on behalf of <person>, agent <agent>." and its record id.

Cost on the card is shown only to the story's owner, its assignee and its budget owner.

06 / AGENT WORK TRACKING

Every write to your tracker is a person's decision

An agent works on the story. People decide what is spent on it and what is written to it.

  • The person approves. By default each comment waits in ADE until the developer approves, edits or discards it.
  • Secrets stay out. ADE scans the comment, and Argent removes anything credential-like before posting.
  • Signed and named. Every write-back is signed in Argent's ledger before it is posted, and the footer names the person and the agent. See One-Click Audit Evidence.
  • The tracker wins. If someone already moved the card in the tracker, Argent leaves it as the tracker has it, and the trail reads REFUSED BY TRACKER.
  • The admin chooses what Argent may write. Per tracker, the summary comment and the move to Done on close are separate switches, and both ship Off.
  • Only a person lifts a hold. No agent tool lifts a hold or approves a comment, and every write an agent asks for waits for its owner's approval.
  • Fix the books, keep the history. Spend booked to the wrong story can be re-tagged to the right one with a reason, and the change is recorded.
07 / AGENT WORK TRACKING

Where it fits and how to start

ForWhy work tracking matters
Financial services and insuranceEngineering teams can show which story each dollar of agent work paid for, and who approved the write.
Retail and healthcareProduct teams cap the stories that matter and see what each story's AI work cost.
Public sectorAgent work on a program is booked to its project, and going past a Pause cap takes a named person's decision.
Your goalsBudget and token spend, audit and accountability, agent identity and ownership, AI visibility, compliance evidence.
Use casesSecuring AI coding agents, AI in regulated and air-gapped environments.
  1. Connect. An admin connects your tracker, and its stories reach ADE.
  2. Cap. Put caps on the stories, epics and projects that matter, set to Pause or Alert only. For company-wide budgets by department, person and agent, see AI Spend Caps.
  3. Write back. Switch on the summary comment and the move to Done on close when your team is ready.
  4. Run it where you need it. Air-gapped and on-premises, private cloud or your public cloud VPC. See all industries and all capabilities.
Argent Strategic Intelligence

Book a demo

See Argent Strategic Intelligence stop a real AI agent, live, with the engineers who built it.