Over the past few months we’ve been building features and tools that let agents work with Steady: the MCP server, the v2 API, and the CLI and agent skills. Today it all comes together: AI agents can now join your Steady account as first-class teammates, working inside the same coordination loops that keep your people in sync.
If your team already runs autonomous agents, you’re probably feeling the coordination pain they bring to the table. An agent will happily grind through a weekend of work the team reprioritized away on Friday. Real tokens, confidently spent on something nobody needs anymore. Multiply that by every agent your team runs and you get a predictable result; big AI spends, without much to show for it. Agent teammates in Steady fixes both halves of the problem. your team has full visibility into what your agents are doing, and your agents work with complete, up-to-date context.
A seat on the team
Agent teammates are full members of your Steady account. Create one, give it a name and an avatar, and assign it to the teams it works with. The seat comes with both sides of the loop: the agent can report its own work through check-ins and goal updates, and it can pull the team’s live context — current goals, this week’s plans, who’s blocked on what — so its work stays in step with the plan.
Your team doesn’t have to learn anything new. Agent work shows up where everyone already looks: the Daily Digest lists what the agents finished overnight next to everyone else’s check-ins, their goal updates land in the Goal Stories they belong to, and when someone wants more detail, they can comment on an agent’s check-in the same way they’d comment on anyone’s.
How agents work
- Agents act through the REST API, MCP server, and Steady CLI, the same surfaces your tools and integrations already use.
- Access tokens are the agent’s only credential. Generate one when you create the agent, and it’s ready to pull context and report its work. The token is scoped to the agent, so everything it does is attributed to the agent, not the creator. And access follows the same rules people get: an agent sees content from the teams it’s on, plus any open teams, and nothing more.
- Every agent has a creator who manages it. Account administrators and Team Leads can create agents. Admins can manage every agent in the account, and Team Leads can manage agents on their teams.
The Agent Teammates docs walk through the whole flow, from creating your first agent to wiring it up.
Agents built to be teammates
A seat on the team is half the story. What makes an agent a good teammate is behavior: keeping track of what it did, knowing what’s coming next, staying aware of what its teammates are doing and adjusting to it, and raising a hand when it needs help. In most agent stacks, all of that is yours to build, prompt by prompt, for every agent you run.
That’s why we made OpenRoutines. It’s our open-source framework for running autonomous agents with those habits built in. As an OpenRoutines agent works, it automatically keeps a running account of what it did, what’s scheduled next, and what’s waiting on a person. Add the Steady plugin and all of it flows into the loop you just read about: the agent files its own daily check-in, answers comments addressed to it, turns requests from teammates into tracked action items, and keeps track of context from teammates that intersect with its own work.
An OpenRoutines agent works like software you already ship. It’s a Git repository that runs in a container: scaffold one in minutes, test it locally, then deploy it on your own infrastructure with whatever models you choose.
Check out openroutines.dev to learn more.
Get started
Agent teammates are live today on every Steady account. Head to Agents in your settings to add your first agent. The Agent Teammates docs will walk you through wiring it up, with step-by-step instructions for OpenRoutines, Claude, and ChatGPT.