Your AI agents, on the clock.
Build agents in chat, equip them with skills and integrations, schedule them like jobs. They ship while you sleep — no agencies, no scripts, no glue code.
No credit card · Magic-link sign in · Bring your own Claude
Powering AI workforces at teams you know
Stop running a chatbot. Start running a workforce.
- Pasting prompts into ChatGPT for the same task every week
- One-off scripts that break the moment a tool changes
- Hiring agencies for work an agent could do tonight
- No memory, no schedule, no team — just a chat tab
- Agents you build once and reuse — chat, schedule, repeat
- Plug-and-play skills + MCP integrations from a curated gallery
- Cron-style jobs that keep working when you're offline
- Team workspaces, roles, and an audit trail on every run
Everything an agent needs. Nothing you have to wire up.
Each piece works on its own. Stack them and your agents go from one-shot prompts to a durable team that runs your back office.
Agent Architect
Describe what you need in chat. The Architect asks the right questions, then ships a fully-configured agent — system prompt, model, skills, integrations, schedule.
Skills gallery
Drop-in capabilities like web research, code review, brand voice, or copy QA. Import any skill from the skills.sh ecosystem or build your own and share with the team.
Integrations gallery
MCP-powered connectors for Slack, GitHub, Drive, Gmail, Linear, Sentry, and more. Add a server, agents pick up the tools — no glue code.
Scheduled jobs
Run any agent on cron — every minute, daily, or once a week. Off-hours work, on time.
Team workspaces
Invite teammates, set roles, and keep every run, thread, and credit auditable.
Local connector
Pair the desktop CLI to run agents on your machine with full file + tool access.
Answers, in five lines or fewer.
- What exactly is an agent on Roster?
- An agent is a system prompt + Claude model + a set of skills and integrations. You can run it on demand from chat or schedule it on a cron — and it shows up alongside the rest of your team.
- How do I build a new agent?
- Open a chat with the Architect. Describe what you need, answer a couple of clarifying questions, and the Architect ships a fully-configured agent into your workspace. Edit any agent later the same way.
- What are skills and integrations?
- Skills are reusable behavior packs (research, copy QA, code review, brand voice) — pull from the skills.sh ecosystem or build your own. Integrations are MCP servers (Slack, GitHub, Drive, Gmail, Linear, Sentry, etc.) you attach to any agent.
- Can agents run on a schedule?
- Yes. Every agent has an optional cron — run it every minute, daily, or once a week. Off-hours work, on time, with the full run history saved per team.
- How does pricing work?
- Plans cover the platform — agents, skills, integrations, schedules, and the Kanban board. Bring your own Claude (Anthropic API key or local connector) and model usage stays on you.
- Is my data secure?
- Workspaces are team-scoped and credentials are encrypted at rest. We never train models on your prompts, runs, or attached files.
Stop running prompts. Start running a workforce.
Sign in with your email. Pick a template or chat with the Architect. Your first agent ships before your coffee gets cold.