Start with a starter agent or recurring workflow.
Triptych OS (Agent OS)
Start with a governed agent.
Choose a starter agent or describe a workflow. Agent OS turns it into a launch plan with context, tools, budget, approvals, runtime, receipts, and optional marketplace exposure.
Starter agents
Pick the job first. Controls come with it.
Full catalog
Browse every governed starter agent
- Ecommerce Support Resolution Agent Maps and resolves repetitive ecommerce support tickets with approval-mode controls and receipts.
- Research Radar Agent Tracks competitors and topics on a schedule and produces sourced evidence cards.
- Sales Prospector Agent Finds prospects with buying triggers, enriches contacts, and drafts outreach for approval.
- Developer Codebase Maintenance Agent Repairs GitHub issues with diffs, tests, and a PR approval gate.
- Context Compiler Agent Compiles PDFs and docs into ECF Core evidence units and Agent OS import files.
- Data Analyst Agent Turns raw datasets into sourced analysis and reproducible summaries.
- Investment Analyst Agent Builds sourced investment briefs under budget and approval controls.
- Daily Briefing Agent Compiles a sourced daily briefing from connected tools and feeds.
- Chief of Staff Agent Coordinates recurring operations, follow-ups, and status roll-ups.
- Recruiter Agent Sources and screens candidates with structured, reviewable output.
- Marketplace Seller Agent Runs marketplace listing and seller-growth workflows under owner gates.
Create a deployment from any starter agent, or build a new agent from a workflow description.
Launch path
Preview first, launch privately, then decide if it can sell.
Budget, approval rules, context, memory, and exposure mode.
Get a first-proof plan before spend, runtime, or publication.
Fund runtime only after owner gates and readiness checks.
Inspect run outputs, cost, evidence, approvals, and reconciliation.
Expose API, marketplace listing, or x402 access only after approval.
Agent OS Command Center
Track runs, cost, proof, and pending decisions.
Live values appear in the deployment dashboard after a private launch. This page stays preview-only.
No live operations from this starter page.
Approvals, memory candidates, budget alerts, and listing drafts appear after launch.
Learning
Agents improve under review.
After each run, Agent OS can propose skills, memories, rubrics, or policy updates. Durable memory and public exposure stay owner-gated.
Reusable work patterns from successful runs.
Versioned, scoped, receipt-backed lessons.
Quality checks for outputs before repeated use.
Integrations
Tools are permissioned capabilities.
Each integration is part of the deployment contract: what the agent can read, write, buy, and what requires approval.