Troubleshooting

Yes. A function is a fixed list of steps, and a step can call a model as well as call an operation or work on data it already has. What makes it a function is that the steps are fixed. See Add a function.

Usually not. Most agents are instructions plus the right operations. Write a script only for work that must wait for a person or must happen in order. See Make it follow steps.

Put it in a script phase. A phase's gate says what must be collected before the run moves on, so the invoice agent can't leave its compare phase without a comparison result. An instruction in the prompt is a request. A gate is enforced. See Make it follow steps.

In the Agent Builder, docked on the left of every agent tab, or directly on the tab. Orca is for provisioning things you don't have yet. See Diagnose and improve.

Something is left to the model that shouldn't be. Check, in order: the script does too little, the agent can't reach what it needs, an instruction can be read two ways, or nothing judges the output. See Diagnose and improve.

Not always. For an MCP server or a knowledge base, choose Discover tools on the connection's Overview tab. For an HTTP API, a database, a Google Sheet or StackOne, you add operations on the connection's Operations tab. A new write operation is enabled as soon as it's added, unless an admin turned that off for the org. See Connections and operations and Approve what agents can do.

Read the error. It names the privilege your database login is missing and gives you the GRANT statement to send to whoever owns the database. The login's own privileges are the boundary, so give a connection that must never write a login that can't write. See Connect a system.

The spreadsheet isn't shared with the service account. Share it with the service account's email address in Google Sheets. See Connect a system.

The operations. The full list of what your agents can do to your finance system is the list of operations enabled on that connection. See Connections and operations.

Three things must be true: the operation is enabled on the connection, the agent holds it in its active version, and if it's a write, it has been approved. Check them in that order. See Approve what agents can do.

Yes, once the send operation is enabled. Approval is given once per operation, not per call, and new write operations are enabled automatically unless an admin turned that off. If each individual email needs a person, add an $ask step to the script: it asks in Slack and the run waits for the answer. An agent can never approve an operation. See Approve what agents can do.

On the connection's Operations tab, press Withdraw approval on a write operation or Disable on a read. That's checked on every call, including in conversations already running. See Approve what agents can do.

Look in Govern → Log, or ask the agent docked on Govern → Monitor. No run recorded means nothing started it, so look at the trigger, the schedule or the agent's availability. See See what happened.

It depends why it stopped. A run that is parked, on a phase wait or an $ask step in Slack, resumes when the wait ends or the answer comes. A run started by a schedule or a webhook that follows a script gets another turn later and continues from the phase it was in. A run your own software started through the API continues when you post another message to the same conversation. Anything else that hit a limit (time, steps or tokens) doesn't continue, and re-running starts from the beginning. Read the record first, because a re-run repeats anything that already happened. See Runs.

Runs have limits on time, steps and tokens, because an agent with a bad instruction keeps going instead of failing. See Limits for the numbers.

Read the gate report in the run. It names the condition that didn't hold. Usually the condition is worded vaguely. See Make it follow steps.

No. Activating a version doesn't run the acceptance tests, and a failing test doesn't block it. Run them yourself before you activate, and treat a failure as a reason not to. What does block activation is an untested operation. See the next question, and Test behavior.

The version puts new script text live, or the function calls operations, with no successful call since they last lapsed in this environment. Run the script's or function's test parameters so the calls succeed, or an org admin can choose Go live untested, which is recorded with their name. See Versions, testing and going live.

That the test met its threshold over several runs. By default a model-judged check runs five times and a code check once. See Test behavior.

It reads from it. An acceptance test uses the agent's real read operations, and refuses every write before it leaves Mindset, so a test never registers a supplier query. Test parameters are different: a test parameter set that writes performs those writes for real once an admin gives the go-ahead. See Test behavior and Versions, testing and going live.

Not in place. An accepted test can't be edited. Amend it, which retires it and accepts a replacement, or retire it with a reason. Dry-run new tests before you accept them. See Test behavior.

Yes. A simulated user with a persona and a goal keeps the conversation going until it reaches the goal, loops, or hits a turn limit, and the whole conversation is scored. See Test behavior.

Everything the agent holds, used through the agent. For the invoice agent, that's read access to the finance system through its read operations, plus the ability to register supplier queries. Nothing else on the finance system. See Build an agent.

No. Ticking The Hub makes an agent visible to every member of the org, in that environment. If only some people should have it, change what the agent holds. See Build an agent.

Check that Available is on, The Hub is ticked, a version is active, and they're in the same environment. See Build an agent.

No. Any member can connect their AI client from the Hub and call the agents available there. Org admins can reach every agent, and have a separate server for building and governing. See Use your agents from AI clients and Start with your coding agent.

No. A call can't be canceled once it starts, so the agent carries on in Mindset. Don't call it again, because that starts a second run. See Use your agents from AI clients.

Yes, if an admin turns on personal agents in Settings → Defaults. A personal agent belongs to its member, is never shared, and holds no connections. It can only hand work to agents already available to its owner. See Roll out Mindset.

Govern → Monitor → Cost: total spend, metered calls, average cost per call, average input tokens per call, tokens in (with the share from cache) and tokens out. Break it down by agent, function, model, provider, user or interaction. See See and control cost.

Run an experiment with the model swapped and compare the results. Take the cheapest model that still passes the tests. An experiment changes nothing on the live agent, so switching the model is a separate step. See See and control cost.

Saving an agent change creates a version and activates nothing. That includes schedules: a new or deleted schedule takes effect when its version is activated. Choose Make it active, or activate it on Versions & Availability. Functions and widgets need publishing. Availability is the one change that takes effect when you press Save. See Versions, testing and going live.

Yes. Save it as a version, then pick that version on the agent's Chat tab. Everyone else keeps the active one. Its writes wait until you press Agree writes. See Change something that's live.

Go back to an earlier version. For an agent, choose Activate version N on it. For a function, choose Roll back to this. For a widget, choose Re-activate. Rolling back an agent or a function can be refused like going forward, if it puts untested operations live. Connections keep no versions: an operation is enabled or it isn't. See Change something that's live.

On the shared deployment, your own agents need your org's model key before they can run. Orca and the builders work in the meantime. An org admin adds one in Settings → Model keys. See Use your agents from AI clients.

Whoever its description says. Org agents have no owner field, so put the owner and a backup in the description in the same format every time. See Roll out Mindset.