Start with the issue you are seeing. These practical checks cover workspaces, agents, connections, runs, approvals and credits, with clear routes to product documentation or contact.
Choose the symptom closest to yours. Each section explains what to check and what to do next.
Check the selected account, required permissions and whether a previous connection needs to be renewed.
Check connection steps →Review the execution status, inputs, connection health and any approval that may be blocking the run.
Review failed-run checks →Sensitive or consequential actions may pause so an authorized person can review them before execution.
Understand approval requests →Reconnect with the intended account, review requested permissions and then retry the affected workflow.
Reconnect an account →Learn how the workspace’s shared credit pool, monthly allowance and usage records relate.
Understand credits and usage →Revoke access for the business app, then confirm that workflows depending on it are reviewed.
Review connection access →Use these checks to narrow the cause. Product access and available connections depend on your current onboarding scope.
Access looks incorrect: confirm that you are using the intended organization and workspace, then ask a workspace owner or admin to review your role.
A team member cannot continue: verify their workspace membership and that their role covers the action they are attempting.
Read workspace and team concepts →An agent is not available: verify the role you need in the specialized agent catalog. Catalog pages describe business roles and do not guarantee that every role is active for every workspace.
Configuration is incomplete: review the agent’s inputs, connected apps, boundaries and required approvals before trying a run.
Review agent documentation →A connection fails: confirm that the account is correct, the provider is reachable and the requested permissions are allowed by your organization.
A connection expired: reconnect the business app, review its permission scope and retry only after the connection reports a valid state.
You need to remove access: revoke the connection and review workflows that depend on it before their next run.
Explore business app integration options →A run failed: note the execution status, identify the failed step, check required inputs and connection health, then retry only after correcting the cause.
A run does not start: look for a missing input, inactive connection, pending approval or usage limit.
Keep the run time, agent role and visible error message when you need further help. Never include passwords, access tokens or other secrets.
Review execution and retry concepts →An approval is pending: confirm that the request has reached a person with the appropriate workspace role and review the action’s inputs and consequences.
An approval was denied: update the underlying instruction, data or permission rather than repeatedly submitting the same action.
Understand human-in-the-loop controls →Usage is higher than expected: review recent runs and their task complexity. The product uses a shared workspace credit pool; precise consumption can vary by the work performed.
A usage limit was reached: check the plan allowance and billing period before starting additional work. This page does not assume an undocumented per-agent cost.
Compare plans and monthly credit pools →Access seems too broad: pause the affected workflow, review the connection scope and remove permissions that are not needed for the defined job.
You suspect unauthorized access: revoke the connection through the relevant provider and contact AI Robot Agents with non-secret details about what you observed.
Review permissions and security controls →If these checks do not resolve the issue, use the contact form. Include the affected workspace, agent role, approximate time and visible error message.
Do not send passwords, API keys, access tokens or full confidential records. No response-time or service-level commitment is implied.
Contact AI Robot Agents about an issue →
Use the documentation hub for setup, configuration, permissions, monitoring and usage concepts.
