01
Approve connector writes
Automation connector writes pause when your policy requires approval. External writes require it by default; reads and internal platform writes follow different rules.
Approval conceptsGovernance
Set the rules for your team and agents
Control access, configure connector approvals, and review recorded actions. Set model usage budgets and chat filters for the work your team and agents carry out.
See the rules in the work
Inspect workflow runs, compare answers, and check the tools and sources available to agents.
01 Automations
Follow the run and its decisions
A configured workflow can review an access report and prepare a notification. Inspect its steps, approval decisions, and recorded effects in the run detail.
02 Arena
Compare wording and check the policy
Read two model responses to the same question. Verify each against your policy before using either one.
03 Agents
Choose the tools the work needs
Equip each project agent with explicit tools and credentials. Managed model calls use Tale’s gateway controls; direct subscription access follows a different path.
04 Knowledge
Read the source behind the answer
Keep policies and runbooks in accessible knowledge. When an answer uses them, inspect the source and check whether it supports the result.
What does governance cover?
Choose which connector writes need approval, which tools agents may use, and which content and usage rules apply.
02
Inspect the proposed operation
Review the destination and exact input before approving. Reject an incorrect operation, fix the workflow or input, and start a new run.
Approval concepts03
Read the audit trail
Inspect recorded actions and approval decisions with their actor and time. Use the audit log to investigate what happened.
Audit logs04
Set organization rules
Configure shared instructions, access policies, and agent tool grants. Enforced permissions and filters complement the instructions you give a model.
Guardrails05
Configure chat content filters
Enable content safety, personal-data detection, and optional moderation for the chat directions you choose. Test the patterns and failure behavior your work needs.
Guardrails06
Set budgets for metered work
Track recorded model usage and set budget rules for supported paths, including managed agent runs. Direct subscription calls use the provider’s own controls.
Governance FAQ
Governance combines access rules, connector approval policies, recorded audit events, usage budgets, and configurable chat filters. Each control has its own scope.
Open the waiting automation run. If it works on a task, the approval card also appears on that task. Connector approvals are distinct from reviewing an agent’s completed task.
Yes. The policy can be configured for a connector or a specific action. External writes require approval by default; changing the policy does not resolve an approval already waiting.
No. The connector broker exposes read actions to agents; governed connector writes run through automation nodes. Direct GitHub tooling and explicitly granted secrets have separate access paths. Review those grants too.
Chat filters can check configured input and output. Model API endpoints filter input only; they do not filter model output. Personal-data detection depends on configured patterns. Assess deployment, providers, and operating procedures together.
Work you can try with your team
Start with a clear brief, defined deliverables, and a review process.
See your team’s next workflow in Tale
Bring a task to a guided demo, or start a workspace and try it with your own agents.