Work orchestration for AI agents

Build pipelines that triage tickets, process documents, and automate research. All from your terminal.

# Triage incoming support tickets $ cat tickets.json | ayo @triager "Categorize by urgency" Processing 23 tickets... ✓ Critical: 2 → escalated to @responder ✓ High: 5 → queued for review ✓ Normal: 16 → auto-responded # Check pipeline status $ ayo flow status @triager → @responder → @closer ↳ 2 tickets in progress

What It Does

A framework for AI agents that work together.

Specialized Agents

Each agent masters one domain. A triager triages. A reviewer reviews. Expertise through focus.

Composable Pipelines

Chain agents with Unix pipes. JSON in, JSON out. Build complex workflows from simple parts.

Local-First Control

Agents live as folders on your machine. Version them in git. Deploy anywhere.

Transform & Route

Each agent transforms data and routes it forward. Filter, enrich, classify, hand off.

Human-Readable State

Tickets, memory, and config are plain Markdown. Read them. Edit them. Understand them.

Terminal-Native

Orchestrate from your shell. Cron jobs, pipes, scripts. No new tools to learn.


Built for Real Work

ayo orchestrates work that doesn't belong in a code editor.

Support Ticket Triage

Route incoming tickets to specialized agents based on content.

$ cat zendesk-export.json | ayo @triager | ayo @router
12 tickets → @billing-agent
8 tickets → @technical-agent
3 tickets → @escalation-agent
Document Processing

Extract and summarize information from bulk documents.

$ find contracts/ -name "*.pdf" | ayo @extractor | ayo @summarizer
Processing 47 contracts...
Extracted key terms → contracts-summary.json
Research Synthesis

Gather information from multiple sources and generate reports.

$ ayo @researcher "competitor pricing Q1 2026" | ayo @analyst
Gathering from 12 sources...
Report generated → research/pricing-analysis.md
Content Review Pipeline

Review drafts and route based on quality assessment.

$ ayo @reviewer drafts/*.md --output approved/
Reviewing 8 drafts...
5 approved → approved/
3 need revision → tickets created
Automated Reporting

Schedule and run recurring reports with flows.

$ ayo flow run weekly-metrics
Collecting data from 4 sources...
Report sent to #metrics-channel

What People Say

Teams using ayo for work orchestration.

"ayo replaced our entire triage spreadsheet workflow. Support tickets now route themselves."
Support Team Lead
Enterprise SaaS
"We process 200+ documents daily now. What took a team of 3 is handled by a single flow."
Operations Manager
Legal Tech
"The Unix philosophy for AI makes so much sense. Agents as directories, pipes between them—it just clicks."
Platform Engineer
Startup

What Makes ayo Different

Not just another chatbot. A framework for AI that actually works.

Agents as Directories

Each agent is a directory with config.json and system.md. Version control them. Share them. Understand them at a glance.

Ticket Coordination

Tickets track work between agents. Add dependencies. Assign to specialists. When one completes, blocked tasks automatically unblock.

Isolated Sandboxes

Agents execute in containers with explicit trust levels. Grant filesystem access, network access, or keep them locked down.

Persistent Memory

Facts survive sessions. Store organizational knowledge. Agents automatically retrieve relevant context when they need it.

Declarative Flows

Define multi-step workflows in YAML. Parallel execution where possible. Automatic dependency resolution. Error handling built in.

Unix Composability

Pipe output between agents. JSON as the universal interface. Script everything. Integrate with your existing tools.


Real Workflows, Not Chatbots

ayo orchestrates AI for actual work—support triage, document processing, research, and more.

Customer Support Triage

A squad of agents categorizes incoming messages, drafts responses, and escalates complex issues—automatically.

$ ayo squad create support-team \ -a @triager,@responder,@escalator $ cat inbox.json | ayo @triager "Process" Created 15 tickets → 12 assigned to @responder → 3 assigned to @escalator

Document Processing Pipeline

A flow classifies documents, extracts key data, validates accuracy, and files the results—with each step handled by a specialist.

$ ayo flow run document-pipeline \ '{"file": "invoice.pdf"}' ✓ classify: invoice ✓ extract: vendor, amount, date ✓ validate: passed ✓ file: /accounting/2024/q1/

Automated Daily Digest

A trigger runs every evening—gathering updates, synthesizing insights, and distributing a summary to your team.

$ ayo trigger create daily-digest \ --cron "0 18 * * 1-5" \ --flow team-digest Trigger created: daily-digest Next run: Mon 6:00 PM

Works With Everything

Connect ayo to your existing tools and data sources.

AI Providers
Anthropic OpenAI Ollama Azure Gemini
Input Sources
JSON CSV PDF Email APIs PostgreSQL S3
Output Targets
Slack Discord Email Webhooks Files
Tools
Crush Browser Shell

Get Started in 30 Seconds

# Install via Homebrew $ brew install ayo-ooo/tap/ayo # Start the onboarding wizard $ ayo onboard
# Install via Go $ go install github.com/ayo-ooo/ayo@latest # Start the onboarding wizard $ ayo onboard
# Clone and build $ git clone https://github.com/ayo-ooo/ayo.git $ cd ayo && go build -o ayo . $ ./ayo onboard
# Run via Docker $ docker run -it ayo-ooo/ayo onboard
Read the Guide

Stay in the Loop

Updates on new features, skills, and workflows. No spam.