What did this mission cost?
Every model call is attributed to its mission, agent and step, with its tokens and cost. Spend stops being one account total on an invoice.
Tracston · Interactive lab
AFPM: AI Flow Performance Monitoring Your AI is working. Do you know what it’s costing you?
Follow the spend from mission to agent, model, and tool. Find repeated work, understand its cause, and explore controls that reduce waste while preserving results.
No sign-up. A fictional mission runs in your browser; nothing touches real systems.
The animation beside this text shows a fictional mission as a 3D topology: a person submits a request to a Coordinator Agent, which branches the work into Research, Pricing and Validation agents that call a model endpoint, a browser, a CRM and an inventory API. The Pricing Agent repeats the search and datasheet summaries the Web Research agent already ran, and every model call carries a large context, so the simulated cost keeps climbing. Imperium captures the Pricing Agent’s model request with its token usage and cost; Observatory correlates it with the earlier calls and traces the cost back to the repeated work. The same mission can be investigated step by step in the lab below, in the “Where did the AI budget go?” scenario.
TRACSTON / AFPM LAB Interactive simulation
Investigate AI agents with Imperium + Observatory.
Choose an investigation
Simulation · fictional data
DelegationTool callModel callOutputPolicy decisionApprovalCorrelation (not a call)Failure / timeout
Mission metrics at the current moment
Simulated figures0 agents0 tool calls$0.000 model cost0 issues
Event list
The same events the graph draws, as text. Only events that have happened by the current time are listed. Keyboard: Space plays or pauses, comma and period step between events, F fits the graph, R resets the view, Esc clears the selection.
| Time | Event | Source | Actor | What | Status |
|---|
Imperium captures and governs.Observatory correlates and explains.
Cost accountability
From AI spend to accountable work.
What caused unnecessary spending?
Repeated searches, duplicate summaries and oversized context show up as patterns across calls. Each finding cites the events behind it, so the cause can be checked.
Did the output justify the cost?
Cost is read next to what the mission delivered and whether it was approved. Controls that cut waste can then be compared against the same result.
Real product screenshots
The products behind the simulation.
These captures of the actual Imperium and Observatory interfaces substantiate the capabilities above: attributing spend, finding its cause and judging the result. Unlike the lab, they show the real products, with demonstration data.
How it works
Imperium captures. Observatory explains.
In Tracston’s AFPM solution, Imperium captures AI activity and its governance context. Observatory uses that telemetry to correlate events, investigate execution, and explain performance, cost, and outcomes.
Step 1 · Capture
Imperium captures and governs
Imperium records each AI request as it happens.
- Who started the mission, and whether that is proven
- Which agent acted, and on behalf of which parent
- Which tool or model it asked for
- Which policy applied, what it decided, and whether enforcement was confirmed
Step 2 · Explain
Observatory correlates and explains
Observatory receives that telemetry and joins the events into the mission and its steps.
- Timing and dependencies along the critical path
- Retry and failure patterns
- Cost by agent and step
- Whether the result met the mission’s requirements
What this means in practice
Observatory is not in the execution path
Agents call their tools and models directly. Observatory receives telemetry about that work. It does not run the mission or route its calls.
The AI detail depends on Imperium
The AI investigation shown here relies on the telemetry Imperium supplies. Without it, Observatory still sees infrastructure signals, such as an HTTP 504, but not which agent, mission or policy they belong to.
Try the AI telemetry switch in the lab to compare.
Scoped to AI flows
This page covers the AI side of AFPM. Observatory’s broader infrastructure monitoring is a separate capability, described on the Observatory page.
Investigation examples
Four questions every AI operator ends up asking.
Each one is a scenario in the lab above. Every scenario can be replayed with the AI telemetry switch off, to see what infrastructure monitoring alone would conclude.
“Where did the AI budget go?”
Where did the AI budget go?
Two agents research the same thing, and the model calls carry about 48,000 tokens of context the steps never use. All costs in the lab are simulated.
- Infrastructure signals alone
- Every model request returns 200 OK. Spend shows only as one total.
- With Imperium telemetry
- Tokens and cost attributed to each agent and step, showing repeated work and oversized context.
Look for input tokens per call, duplicate searches and summaries, and cost by agent.
“Why is this taking so long?”
The slow mission
The Pricing Agent keeps calling an inventory endpoint that times out, and retries without a fallback. The whole quotation waits on it.
- Infrastructure signals alone
- Four HTTP 504 responses from inventory-api. No caller, no mission, no delay.
- With Imperium telemetry
- Which agent was retrying, for which mission, and that the retries sat on the mission’s critical path.
Look for consecutive timeouts from one agent to one tool, and the branch on the critical path.
“Who requested this data?”
Sensitive data
A child agent asks the CRM for restricted fields. The policy decision is DENY, but no enforcement is recorded and the CRM answers. The mission came from a shared service account, so the person behind it is not proven, and the lab says so.
- Infrastructure signals alone
- A normal CRM read that returned 200 OK.
- With Imperium telemetry
- The request, the identity evidence, the policy decision and the enforcement record, as four separate facts.
Look for a decision without enforcement evidence, and an initiator that stays unproven.
“Did the mission actually succeed?”
False success
Every agent finishes and validation passes, but validation never checked the inventory confirmation. The approver rejects the draft. Task completion, technical validation and the business outcome are three different measurements.
- Infrastructure signals alone
- All services healthy. Nothing is flagged.
- With Imperium telemetry
- What validation checked and skipped, what the inventory call actually returned, and the approver’s decision.
Look for what validation checked, what the tool actually returned, and the business decision.
Four distinctions the lab keeps
- Retry loop
- An agent repeats a failing call without changing its approach. Each attempt looks reasonable; together they stall the mission.
- Denied is not enforced
- A policy decision records what should happen. Enforcement evidence shows what did happen. An investigation needs both.
- Unknown identity
- When a mission starts from a shared account with no user session, the honest answer to “who asked?” is “not proven”.
- Completed is not successful
- Every step can return OK while the output misses a requirement. Success is measured against the mission, not the steps.
Understand your AI spend
Explore a demonstration of Imperium and Observatory working together.