Portable AI workflows in
simple YAML

Orchestrate LLM agents with built-in observability, cost tracking, and security.

Review PRs for security issues. Run via CLI, GitHub webhook, or CI pipeline.

name: security-review
description: Review PRs for security issues

inputs:
  - name: owner
    type: string
  - name: repo
    type: string
  - name: pr_number
    type: number

steps:
  - id: get_diff
    github.get_pull_request_diff:
      owner: "{{.inputs.owner}}"
      repo: "{{.inputs.repo}}"
      number: "{{.inputs.pr_number}}"

  - id: review
    type: llm
    model: balanced
    prompt: |
      Review this diff for security vulnerabilities.
      Flag any issues found with severity and remediation.

      {{.steps.get_diff.content}}

  - id: comment
    github.create_comment:
      body: "{{.steps.review.response}}"

name: iac-review
description: Parallel analysis of infrastructure changes

steps:
  - id: plan
    shell.run: terraform plan -no-color
    workdir: ./infrastructure

  - id: analyze
    type: parallel
    steps:
      - id: risk
        type: llm
        model: balanced
        prompt: |
          Summarize the RISK of these infrastructure changes.
          Flag any security concerns, cost increases, or breaking changes.

          {{.steps.plan.stdout}}

      - id: network
        type: llm
        model: balanced
        prompt: |
          Summarize the NETWORK OPERATIONS impact of these changes.
          Note any IP changes, security group updates, or routing modifications.

          {{.steps.plan.stdout}}

  - id: report
    type: llm
    model: fast
    prompt: |
      Create a brief executive summary combining these analyses:

      ## Risk Assessment
      {{.steps.analyze.risk.response}}

      ## Network Impact
      {{.steps.analyze.network.response}}

name: incident-response
description: Triage alerts and suggest remediation

inputs:
  - name: service
    type: string
  - name: message
    type: string

steps:
  - id: triage
    type: parallel
    steps:
      - id: logs
        shell.run: kubectl logs -l app={{.inputs.service}} --tail=50

      - id: changes
        github.list_commits:
          repo: "{{.inputs.service}}"
          since: "24h"

  - id: analyze
    type: llm
    model: balanced
    prompt: |
      Alert for {{.inputs.service}}: {{.inputs.message}}

      Recent logs:
      {{.steps.triage.logs.stdout}}

      Recent commits:
      {{.steps.triage.changes.commits}}

      Provide: 1) Root cause hypothesis  2) Suggested fix

  - id: notify
    slack.post_message:
      channel: "#incidents"
      text: "🚨 *{{.inputs.service}}*\n{{.steps.analyze.response}}"

name: oncall-handoff
description: Generate shift handoff summary

steps:
  - id: gather
    type: parallel
    steps:
      - id: incidents
        pagerduty.list_incidents:
          status: open

      - id: deploys
        github.list_deployments:
          environment: production
          since: "12h"

      - id: issues
        jira.search:
          jql: "priority = Critical AND status != Done"

  - id: summarize
    type: llm
    model: balanced
    prompt: |
      Create a concise on-call handoff:

      Open incidents: {{.steps.gather.incidents}}
      Recent deploys: {{.steps.gather.deploys}}
      Critical issues: {{.steps.gather.issues}}

      Format: What's hot, what to watch, pending actions.

  - id: post
    slack.post_message:
      channel: "#oncall"
      text: "📋 *Shift Handoff*\n{{.steps.summarize.response}}"

Everything you need for production AI workflows

YAML-First

Define complex workflows in simple, readable YAML. No SDK required.

Any LLM

Anthropic, OpenAI, Ollama, or others with one config change.

Production Ready

Built-in observability, cost tracking, and security controls.

Flexible Runtime

CLI, API server, webhooks, or scheduled execution.

Connectors

GitHub, Slack, Jira integrations with zero code.

Auditable

Security review YAML, not agent code scattered across codebases.

Real-world use cases

See how teams use Conductor to automate their workflows.

Code Review Bot

Automated PR reviews with multi-persona feedback.

Issue Triage

Classify and route issues to the right team automatically.

Build Failure Analysis

Diagnose CI failures and suggest fixes.

How it works

Write YAML

Define your workflow in a simple YAML file

Configure

Set your LLM provider credentials

Run

Execute from CLI, API, or on a schedule

Iterate

Refine prompts and add more steps

Ready to get started?

Install Conductor and run your first workflow in minutes.

go install github.com/tombee/conductor@latest

Read the Docs

Portable AI workflows in simple YAML