An agent for every role on the team.Run them as shipped, or build your own.
The library covers the delivery lifecycle role by role, from backlog readiness to release evidence. Every agent is a versioned definition that runs inside your Azure DevOps project, reads only the scope you configure, and pauses for a named approver wherever your policy says so.
Agents for
everything DevOps
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.
22 roles in six groups.
Pick the group that matches your team. Each role lists the workflows it can hand over, from Definition of Ready checks to evidence packages, and the seven featured agents have full pages with example runs.
Product & requirementsBusiness Analyst Agent
Checks work-item quality, drafts and validates acceptance criteria, decomposes epics, asks for missing information and keeps the links downstream teams depend on.
Definition of ReadyAcceptance criteriaEpic decompositionRequirement change impactAgent page →
Product & requirementsProduct Owner Agent
Assesses backlog readiness, prepares acceptance criteria, splits epics and supports prioritization with your configured value, risk, dependency and industry factors.
Backlog readinessPrioritization supportEpic and feature decompositionRelease-scope reviewAgent page →- Product & requirements
Requirements Engineer Agent
Prepares formal specifications, classifies requirement levels, links needs to system and software requirements, maps them to standards and analyzes change impact against the baseline.
Formal requirement authoringUpstream and downstream linkageStandards mappingRequirements baselineAgent page → - Product & requirements
UX Designer Agent
Turns experience work into testable delivery criteria: UX acceptance criteria, accessibility checks, missing-state questions and design feedback triage on the work item.
UX acceptance criteriaAccessibility requirementsMissing-state reviewHandoff readinessSee it in a demo → - Development & architecture
Developer Agent
Reviews pull requests against linked work, connects commits to work items, investigates defects and prepares fixes under the repository policies already in place.
Contextual PR reviewCommit and PR linkageBug investigationReviewer and code-owner routingSee it in a demo → - Development & architecture
Software Architect Agent
Drafts architecture decision records, checks design-review readiness, analyzes design changes and keeps architecture documentation linked to the requirements it serves.
Architecture decision recordDesign-review readinessDesign change impactRequirement-to-design linkageSee it in a demo → - Development & architecture
Systems Architect Agent
Decomposes systems into subsystems, allocates requirements, keeps architecture views and interfaces consistent and traces system changes through to verification.
Subsystem decompositionRequirement allocationModel consistencySystem change impactSee it in a demo → 
Delivery & operationsDevOps Engineer Agent
Diagnoses failed pipelines, compares runs, audits build configuration, prepares CI/CD and infrastructure changes for approval and verifies pipeline health over time.
Pipeline failure analysisPipeline remediationDeployment readinessPipeline healthAgent page →- Delivery & operations
SRE / Platform Engineer Agent
Correlates alerts with recent deployments, commits and runbooks, prepares or runs approved remediation, verifies recovery and opens traceable follow-up work.
Incident investigationChange correlationRecovery verificationRunbook maintenanceSee it in a demo → 
Delivery & operationsRelease Manager Agent
Checks exit criteria, verifies change approvals, drafts release notes and rollback plans, and assembles a review-ready package for the go/no-go decision.
Readiness verificationRelease notesRollback planRelease packageAgent page →- Delivery & operations
Configuration Manager Agent
Captures and compares baselines, tracks engineering changes, reconciles as-built against as-designed records and prepares configuration status reports.
Configuration baselineEngineering change orderAs-built versus as-designedStatus accountingSee it in a demo → 
Quality & validationQA Engineer Agent
Triages defects, generates test cases, finds coverage gaps, investigates failed and flaky tests and prepares quality reports with the evidence linked.
Bug triageTest generationCoverage analysisFlaky-test diagnosisAgent page →- Quality & validation
Test Engineer Agent
Designs verification and validation tests, maintains test configurations, coordinates runs, investigates failures and produces evidence aligned to each requirement and gate.
Verification test designValidation test designFailure investigationRetest and regressionSee it in a demo → - Quality & validation
Acceptance Coordinator Agent
Checks stage-gate readiness, assembles review packs, verifies acceptance-test coverage and keeps gate findings and approval history in one place.
Stage-gate readinessReview packAcceptance coverageAcceptance reportSee it in a demo → - Leadership & coordination
Engineering Manager Agent
Builds team status, flow and cycle-time analysis, blocker summaries and workload signals from the work already recorded in Azure DevOps.
Weekly delivery statusFlow and cycle-time analysisBlocker and risk reviewDashboard maintenanceSee it in a demo → - Leadership & coordination
Project Manager Agent
Tracks milestone health and critical paths, detects dependencies, flags program risk and assembles status reports and gate packs from the underlying work.
Milestone healthCritical-path analysisDependency detectionGate review packSee it in a demo → - Leadership & coordination
Scrum Master Agent
Prepares sprint plans and reports, tracks impediments, maintains the Definition of Done and surfaces stale work before the team's ceremonies.
Sprint planning preparationImpediment trackingRetrospective preparationStale-work detectionSee it in a demo → - Leadership & coordination
Technical Writer Agent
Keeps product docs, wiki pages and runbooks in step with approved changes, and preserves links back to the source artifacts each page describes.
Documentation generationWiki synchronizationRunbook updateRelease documentationSee it in a demo → - Leadership & coordination
Team Assistant Agent
Classifies and routes incoming work, builds queries, links related items, summarizes long discussions and flags stale or duplicate work across the team.
Classify and route workCustom query creationDiscussion summaryStale and duplicate workSee it in a demo → - Risk, security & compliance
Security Engineer Agent
Triages findings, investigates dependency vulnerabilities, verifies fixes, keeps third-party inventories current and opens remediation work with the evidence attached.
Finding triageDependency vulnerability investigationFix verificationRemediation workflowSee it in a demo → - Risk, security & compliance
Risk Manager Agent
Maintains risk or hazard registers, defines controls, traces risks to requirements and verifies implementation evidence using your approved analysis methods.
Risk or hazard registerControl definitionRisk-to-requirement traceabilityResidual riskSee it in a demo → 
Risk, security & complianceCompliance Officer Agent
Audits lifecycle traceability, finds compliance gaps, checks approval trails and assembles source-linked evidence packages for qualified review.
Traceability auditCompliance-gap analysisElectronic approval trailEvidence packageAgent page →
Role pages organize workflows and vocabulary. They are not employee-replacement pages, and no agent holds professional accountability. The role owner keeps judgment, approval and regulated determinations: agents prepare, validate, trace and assemble review-ready outputs, and qualified people decide.
The exact agent you need, in plain language.
Start from a library agent or a blank definition. No coding is needed: the platform generates what runs behind the instructions, and you never write or see that code.
Start from the library, or describe it.
Add a library agent with Use, or choose Create with AI and describe the job in plain English. The platform drafts the agent and matches it to skills from the library, and you review the result before anything is saved.
Give it your rules and its scope.
Edit the instructions in plain English, attach your SOPs, templates and standards as knowledge, choose the model and see its credit rate, and scope the agent to a project, area path or work item type.
Align it to your process template.
Library agents are built on predefined process templates. Compatibility Alignment maps their work item types, fields and allowed values to yours, suggests matches, flags type mismatches, and blocks invalid mappings until every conflict is resolved.
Set the trigger and save a version.
Choose a manual, event or scheduled trigger and decide whether Required Approval is on. Every save is a version you can preview and restore. When a library template changes, you choose Keep Existing or Review & update, and the previous version stays in Version history.
Match the agent to the shape of the work.
Execution Agents run a fixed rule your team can read. Assisted Autonomous Agents reason over context and draft for review. Most teams use both, and every plan includes both.
| Execution Agents | Assisted Autonomous Agents | |
|---|---|---|
| How it decides | Applies a fixed, versioned rule. Deterministic code, the same result for the same input. | Plans with model-based reasoning inside the configured tool and data scope. |
| Model at run time | None. No model call when the agent runs. | A model call on every run, on the model you chose for that agent. |
| Cost per run | 4 credits flat, about $0.04, however many Azure DevOps calls the rule makes. | Scales with the work: invocation, model tokens, tool calls and orchestration steps. |
| Writes to Azure DevOps | Updates fields, tags, states, links or scores directly, because the rule is fixed. | Leaves a draft or analysis on the item, and writes back once the reviewer approves. |
| Approval | The approval is the rule itself: your team wrote it, can read it, and versions it. | With Required Approval on, the run pauses and opens a task for the named approver. |
| Best suited to | High-frequency checks: required fields, linkage, readiness flags, nightly hygiene sweeps. | Drafting criteria, decomposing epics, pipeline diagnosis, change impact, evidence assembly. |
| How it is created | Generated from a plain-language prompt, with model credits charged once at creation, or written by hand at no model cost. | From a library agent, from a plain-English description, or from instructions you write yourself. |
| Run guardrails | Execution-time and API-call limits. A run that reaches one is stopped. | Maximum steps, retries and sub-agents per run. A run that reaches one is stopped. |
78 persona agents for six more industries.
Each industry is its own section under Assisted Autonomous Agents, with 13 role-based personas such as Product Owner, Business Analyst, Systems Architect, Compliance & Risk Officer and Requirements Engineer, plus engineering and delivery roles specific to the domain. Add one with Use, and Compatibility Alignment maps it to your process template.
- 13 personas
Defence
Capability backlogs under JCIDS, DoDAF 2.02 architecture viewpoints, and NIST SP 800-171, CMMC 2.0 and DFARS cybersecurity compliance.
- 13 personas
Oil & Gas / Energy
Process control and Safety Instrumented Systems under IEC 61508/61511 (ISA-84), IEC 62443 zone and conduit security, and OSHA PSM and API RP 754 compliance.
- 13 personas
Hardware & Electronics
Backlog and requirements rigour aligned with IPC-2221, IPC-6012, IPC-A-610, JEDEC and IEC 62368-1.
- 13 personas
Rail Infrastructure
EN 50126 RAMS, EN 50128 software, EN 50129 signalling safety and EN 50159 safe transmission, covering interlocking, train detection, ETCS/ERTMS and CBTC architecture.
- 13 personas
Construction & Civil Engineering
ISO 19650 information management, BIM federation, and Eurocode, IBC and ASCE 7 compliance.
- 13 personas
Nuclear
Safety-classification-aware design basis, system and software requirements, I&C and plant architecture with independence and single-failure criteria, and Part 21 evaluability screening.
- 150+agents in the library
- 200+core capabilities
- 800+industry-tuned skills
- 15industry packs
[ Trusted by Microsoft and industry leaders ]
Questions about the library.
Execution Agents run deterministic code with no model call at run time, so every run costs a flat 4 credits. Assisted Autonomous Agents use a model on each run to read context, reason and draft, and their cost scales with the work performed. Put rule-shaped, high-frequency work on Execution Agents and spend credits where reasoning adds value.
No. Describe the agent in plain English, write the instructions yourself, or start from a library agent. The platform generates the code behind it. If you prefer to write an Execution Agent's code yourself, you can, and there is no model cost at all.
Yes. Run it as shipped, or fork it: edit the instructions in plain English, attach your SOPs and templates as knowledge, narrow its scope and change its model. Every save is versioned, and you can preview and restore any earlier version.
Your agent is flagged with a Library Update icon and shows the Updated tag in the library. Choose Keep Existing to stay where you are, or Review & update to open the new library content in the editor. If you edited the agent, a warning says the update replaces those edits, and your previous version stays in Version history.
Compatibility Alignment checks the agent's work item types, fields and allowed values against your project when you add it. It maps what matches automatically, suggests matches for the rest, and lets you map each value individually before the agent is saved.
No. Qualified people keep architecture, safety, compliance, certification and go/no-go decisions. Agents prepare, validate, trace and assemble review-ready outputs, and anything with side effects follows the approval policy your team configured.
Yes. Every plan includes both agent types, the full agent catalog and every industry pack. Plans differ in execution capacity and operating capabilities, not content.
Put one agent on one real workflow.
Start the free trial with 2,000 credits and no card, or bring a backlog, a pull request or a failed pipeline to a 30 minute demo and watch the agent run inside Azure DevOps.