Star our project on GitHub

Your code commits.
Our agents handle the rest.

Stop drowning in DevOps busywork. Deploy AI micro-agents that eliminate 80% of repetitive tasks - without touching your existing toolchain.

bash
npm install -g @autodevops/verifier
30% faster PR cycles
Agentic workflow
How It Works

AI agents that live in your existing workflow

No new platforms. No migrations. Just intelligent automation that plugs into the tools you already use.

1

Hook into existing events

Agents trigger on git commits, PR merges, issue creation—wherever work happens.

2

Execute focused micro-tasks

Each agent masters one thing: linting, security scanning, doc generation, test creation.

3

Review and approve results

Agents suggest changes. Engineers stay in control. Build trust incrementally.

yaml
# .autodevops.yml
agents:
  pre-commit:
    - name: polyglot-linter
      auto_fix: true
      languages: ["js", "py", "go"]
    
    - name: security-scanner
      block_on_critical: true
      
  post-merge:
    - name: doc-updater
      targets: ["README", "API_DOCS"]
    
    - name: adr-writer
      template: "decisions/template.md"
      
  on-issue-created:
    - name: spec-generator
      output: "specs/"

How It Works

Visual overview from local hooks to AI agents and CI/CD orchestration.

diagram

Rendering diagram…

Features

Agents that developers actually want to use

Start with the most painful tasks. See immediate value. Scale as trust grows.

Polyglot Linter

Auto-fix style issues across all languages. No more nitpicky PR comments.

Saves 45min per PR cycle

Security Scanner

Catch vulnerabilities and secrets before they hit production.

Prevents 89% of security incidents

Test Generator

Creates test skeletons and edge cases based on your code changes.

3x faster test coverage

Skill Auto-Activation

Detects intent, loads the right Claude skills, and injects guidance before coding begins.

Consistent patterns across 300k+ LOC

Dev Docs Guardrails

Keeps plans, context, and task checklists in sync so agents never lose the plot.

Zero lost planning sessions

PM2 Observability

Managed microservices with automatic log streaming, restarts, and Claude-ready diagnostics.

Debug time cut by 60%

Pre-Commit Analytics

OpenTelemetry-powered dashboards track token costs, issues prevented, and verification performance in real-time.

Complete visibility into AI verification

Open Source CLI

Start automating in minutes

Our open-source CLI brings AI automation to your terminal. Run locally or in CI—your choice, your control.

Zero config setup

Works out of the box with sensible defaults

Language agnostic

JavaScript, Python, Go, Rust—we support them all

Privacy first

Your code never leaves your infrastructure

terminal
bash
# Install globally
npm install -g @autodevops/verifier

# Initialize in your repo
verifier init

# Run your first agent
verifier run lint --auto-fix

# See all available agents
verifier list

# Run in watch mode
verifier watch
Pre-Commit Analytics

Complete visibility into AI-generated code verification

OpenTelemetry-powered observability gives you real-time dashboards for token costs, issues prevented, security findings, and verification performance.

Know exactly what your AI agents are doing

Every pre-commit verification is instrumented with OpenTelemetry traces, metrics, and events. Export to Grafana, Datadog, or any OTel-compatible backend for enterprise-grade observability.

Cost Management

Track token usage and LLM costs per agent, repository, and team. Set budget alerts before you exceed limits.

Issues Prevented

Measure ROI by tracking security vulnerabilities, lint violations, and bugs caught before they reach production.

Performance Insights

Monitor agent execution times, cache hit rates, and API latencies. Optimize your verification pipeline with data.

grafana dashboard

127

Sessions Today

342

Issues Prevented

8

Commits Blocked

$12.47

Cost (24h)

Verification activity over time

AgentTokensCostSuccess
security-scan52K$3.6787%
lint-agent45K$4.2394%
.verifier/config.yaml
yaml
# Enable OpenTelemetry export
telemetry:
  enabled: true
  serviceName: verifier-precommit

  # Export to your observability stack
  otlpEndpoint: http://localhost:4317
  protocol: grpc

  # Or use Grafana Cloud
  # otlpEndpoint: https://otlp.grafana.net

  # Team attribution
  resourceAttributes:
    department: engineering
    team.id: platform
    cost_center: eng-123
metrics exported
yaml
# Metrics streamed to your dashboard
verifier.session.count        # Verification sessions
verifier.agent.tokens         # Token usage by agent
verifier.agent.cost           # Cost per execution
verifier.issues.count         # Issues by severity
verifier.commit.blocked       # Blocked commits
verifier.cache.hits           # Cache performance
verifier.budget.utilization   # Budget tracking

# Events for detailed analysis
verifier.session_start
verifier.agent_result
verifier.security_finding
verifier.commit_blocked

Enterprise Ready

Export to any OTel-compatible backend: Grafana, Datadog, New Relic, Honeycomb, or self-hosted Prometheus.

Team Attribution

Track costs by team, project, or cost center with custom resource attributes. Enable chargeback and usage reporting.

ROI Measurement

Quantify value: issues prevented × estimated fix time = development hours saved. Prove AI verification ROI.

Claude Skills for CI/CD

Supercharge your pipeline with Agent Skills

Package domain expertise as modular Skills. Claude loads exactly what each CI/CD stage needs—no context bloat, unlimited scale.

Progressive Disclosure

Skills load in stages: metadata at startup (~100 tokens), instructions when triggered (<5k tokens), resources on-demand (effectively unlimited).

Install hundreds of Skills with zero context penalty

Only relevant Skills activate in each pipeline stage

Bundle unlimited reference docs without token cost

Filesystem-Based Architecture

Skills are directories with SKILL.md instructions, executable scripts, templates, and docs. Claude accesses via bash—like an onboarding guide.

Scripts execute outside context—results only

Version control Skills like code

Share custom Skills across teams and repos

.claude/skills/security-audit/SKILL.md
yaml
---
name: security-audit
description: Comprehensive security scanning for CI/CD pipelines. Use when validating code before deployment or when security review is required.
---

# Security Audit Skill

## Quick Start
Run the security scanner on staged files:

```bash
bash scripts/scan.sh --staged
```

## Resources
- [CWE_DATABASE.md](CWE_DATABASE.md) - Common weakness enumeration
- [OWASP_TOP10.md](OWASP_TOP10.md) - Security best practices
- [schemas/findings.json](schemas/findings.json) - Output format

## Pipeline Integration
Triggers: pre-deploy, on-PR
Blocks: critical vulnerabilities (CVSS ≥ 9.0)
Outputs: .security-report.json

CI/CD Pipeline Integration

Claude Skills transform generic agents into CI/CD specialists. Each pipeline stage triggers relevant Skills—lint, test, security, deploy—without manual prompting.

1

Pre-commit: Lint & Test Skills

Load code quality Skills before commit

2

Pre-deploy: Security & Compliance Skills

Validate against regulatory requirements

3

Post-merge: Documentation Skills

Auto-generate docs with PowerPoint, Word, PDF Skills

Pre-built Skills

PowerPoint, Excel, Word, PDF Skills work immediately in CI/CD. No setup required.

pptx
xlsx
docx
pdf

Custom Skills

Package your team’s workflows as Skills. Share across repos, version control like code.

SKILL.md
scripts/
templates/

Token Efficiency

Metadata: ~100 tokens. Instructions: <5k. Resources: effectively unlimited via filesystem.

Progressive Loading
No Context Bloat