[ Agent analytics ]
Consumption by agent.
Token and credit use by agent and module, over the period you select.
Agents that execute work across Azure DevOps.
AI that acts. Not just assists.
No card required. Your data stays in your Azure DevOps tenant.
[ Trusted by Microsoft and industry leaders ]
[ The catalog ]Agents for
Agents for
everything DevOps
Plan
Business Analyst Agent
Drafts acceptance criteria, decomposes epics, and keeps the backlog complete before refinement starts.
OwnProduct Owner Agent
Builds sprint proposals, release notes, and acceptance packets, so decisions start from assembled facts.
TestQA Engineer Agent
Drafts test cases from criteria, maps coverage gaps, and writes triage-ready defects from failed runs.
ShipDevOps Engineer Agent
Triages pipeline failures with evidence, reviews YAML changes, and audits environment checks nightly.
ReleaseRelease Manager Agent
Sweeps release scope for open criteria and pending approvals, and assembles the rollback pack before go/no-go.
GovernCompliance Officer Agent
Walks trace chains, audits branch policies nightly, and freezes audit evidence packs on demand.
TraceRequirements Engineer Agent
Runs suspect-link analysis, verifies baselines, and posts impact summaries before a change is called controlled.
Create with AICreate your own agent
Describe it in plain English. The platform generates the agent, matched to skills from the library. You never see code.
[ Orchestration ]
One request spins up an entire team.
Ask for release readiness and the work fans out: one agent sweeps the criteria, one maps test coverage, one assembles the approval trail. Explicit handoffs, one checkpoint before anything is written back.
Agents at workRUNS TODAY
Drafted acceptance criteria on STORY 4821
Business Analyst Agent · event: moved to Ready
Reviewed PR !412 against linked criteria
Product Engineer Agent · event: PR opened
Pipeline fix proposed, awaiting approval
DevOps Engineer Agent · checkpoint held
Generated 6 test cases from changed criteria
QA Engineer Agent · event: criteria updated
Nightly branch-policy audit completed clean
Compliance Officer Agent · schedule: 02:00
Release notes drafted, waiting on sign-off
Product Owner Agent · checkpoint held
Evidence pack assembled for release 2.4
Compliance Officer Agent · manual run
Backlog hygiene sweep · 12 items annotated
Business Analyst Agent · schedule: weekly
ILLUSTRATIVE FEED · EVERY RUN LANDS IN YOUR JOB LOG
Not a prompt. A controlled run.
Pick any workflow. The mechanics never change: trigger, context, work, checkpoint, write-back, record.
- WORKFLOW 01
Draft acceptance criteria.
A story moves to Ready. The agent reads the item, its links, and the discussion, then drafts Given-When-Then criteria with edge cases, and waits for the BA to approve.
- WORKFLOW 02
Review a pull request.
A PR opens against main. The agent checks it against the linked acceptance criteria, flags the gap, and posts inline comments, before a human reviewer spends a minute.
- WORKFLOW 03
Triage a failed pipeline.
The nightly build goes red. The agent correlates the failure with merges since last green and flaky history, classifies it with evidence, and proposes the fix for approval.
- WORKFLOW 04
Assemble audit evidence.
An auditor asks for the trace. The agent walks requirement to test to result across the scope, flags two broken links, and exports the evidence pack for review.
dev.azure.com/contoso · Boards
run started · STORY 4821 → Ready
reading item, 6 links, discussion history
drafting criteria · 4 scenarios, 1 edge case
Given a guest user with items in the cart…
checkpoint · draft on item, awaiting BA approval
write-back only after sign-off
150+ agents in the library
Explore all agents- 200+
- core capabilities
- 800+
- industry-tuned skills
- 15
- industry packs
AI that acts. Not just assists.
Anatomy of a run.
Seven stages, every run, in order. Click a stage or let it play.
01 Trigger
A user request, an Azure DevOps event, or a schedule starts the run. Nothing fires that your team did not configure, and a disabled agent does not execute at all.
Knows your work. Keeps your rules.
Capabilities
Memory and knowledge
Attach your SOPs, templates, and standards as knowledge. The agent keeps how your team works and applies it on every run.
criteria_format "Given / When / Then"standards "ISO 26262 · GAMP 5"templates "release notes v3"scope "payments · area: checkout"naming "REQ- / HAZ- / TC-"approver "named, per workflow"KNOWLEDGE UPDATED · SOP v3 attached
Knowledge lives with the agent definition, versioned like everything else. Update a template once and every future run uses it. Nothing is learned silently: what the agent knows is what your team attached, and you can read all of it.
Run it like infrastructure.
The operational surfaces your platform team will ask about, straight from the product.
[ Top agents · runs ]
Runs verified on first pass
0%
ILLUSTRATIVE DATA · EVERY RUN LANDS IN YOUR JOB LOG
Top agents · runs
- 1.
Business Analyst123 - 2.
QA Engineer98 - 3.
DevOps Engineer87 - 4.
Compliance Officer73 - 5.
Release Manager62 - 6.Requirements Engineer61
[ Credit status ]
Used / remaining.
Daily alerts at your threshold.
[ Usage dashboard ]
- triggerwork item 4821 updated
- checkedcriteria · links · fields
- changed+1 follow-up task
- approvalawaited → granted
- outputdraft accepted · verified
[ Live runs ]
Watch agents work.
Every trigger, checkpoint and result, as it happens.
Detected
- scanning…
[ Native to Azure DevOps ]
No separate portal, no sync, no new login.
Installs from the marketplace into your Azure DevOps organization. Outputs land on the work item, the PR, or the wiki page where the team already looks.
FIG 1
FIG 2
FIG 3
Every run, reconstructable.
What triggered it, what it checked, what it changed, and the output it produced.
Versioned, with restore.
Preview and restore in one step.
Allow and deny, per group.
Set at project level with organization inheritance.
Your model, your rate.
See its credit rate at selection.
Fires on your events.
An agent cannot add its own triggers.
Verified, then recorded.
A run is not treated as successful until the result is verified.
Permission is explicit. Approval is named.
Your data stays in Azure DevOps
Agents run against your tenant on the same security and trust model as Copilot4DevOps, with no separate portal and no new login.
Approval before anything ships
Assisted autonomous drafts pause for a named approver. Execution agents act only on fixed rules your team wrote and can read.
- ✓
Evidence pack assembled for release 2.4
14:06:06 · Compliance Officer Agent
- !
Release notes drafted, waiting on sign-off
14:05:19 · Product Owner Agent
- ✓
Nightly branch-policy audit completed clean
14:04:32 · Compliance Officer Agent
- ✕
Draft rejected · nothing written to the project
14:03:45 · Product Owner Agent
- !
Pipeline fix proposed, awaiting approval
14:02:58 · DevOps Engineer Agent
- ✓
Drafted acceptance criteria on STORY 4821
14:02:11 · Business Analyst Agent
Never trained on your data
Your project data is not used to train models. Enterprise plans bring your own LLM and key for full data-path control.
Verified, then recorded.
Post-action checks confirm the expected result exists, and the job log keeps the whole run.
[ Cost design ]
When rules run, the meter barely moves.
Execution Agents use a language model once, at creation. At run time they execute deterministic code with no model call, so the checks that fire on every work item update or nightly sweep cost near zero.
Credits are pooled across your organization with no per-user fees, agents pause cleanly at zero, and every plan includes both agent types and all industry packs.
See pricingRun cost
Credits · pooled across your organization
$0.01/credit
No per-user fees · pause cleanly at zero
Common questions.
No. You describe the agent in plain English, write the instructions yourself, or start from a library agent. The platform generates the code behind it, so you never write or see it.
No. Agents run inside your Azure DevOps environment, on the same security and trust model as Copilot4DevOps, with no separate login and no data exported to a third-party system.
Execution agents apply a fixed rule you wrote, with no AI involved at the moment of execution. Assisted autonomous agents draft or analyze, and pause for your approval according to the policy you configure.
Both. Execution agents can update fields, tags, or scores directly, since the rule is fixed. Assisted autonomous agents leave a draft for review and update the item only once approved.
Plans start at $500 a month with 50,000 credits, pooled across the organization with no per-user fees. Execution agent runs cost a flat 4 credits, the free trial includes 2,000 credits, and agents pause cleanly when credits reach zero.
Put an agent to work this week.
Free trial with 2,000 credits, no card. Or bring one workflow to a 30 minute demo and watch the run live.