> ## Documentation Index > Fetch the complete documentation index at: https://docs.coverbase.com/llms.txt > Use this file to discover all available pages before exploring further. # Coverbase AI: memory, permissions and standing agents > Review what Coverbase AI remembers from your corrections, choose which changes it may make without asking, and set up agents that run on a schedule or when something happens. For AI agents: a documentation index is available at [https://docs.coverbase.com/llms.txt](https://docs.coverbase.com/llms.txt). This page is also available in markdown by appending .md to the URL. > [!NOTE] > > This guide is part of the [User Guides](https://docs.coverbase.com/user-guides/overview.md) collection. It covers **Configuration > Coverbase AI Memory** and **Configuration > Coverbase AI Agents**. For starting and steering a run, see [Working with Coverbase AI](https://docs.coverbase.com/user-guides/coverbase-ai.md). Coverbase AI learns from the corrections you make, and you decide how much it may do without you. Both live under **Configuration**. The mistake people make most often is correcting the same thing run after run. Say it once as a standing instruction, such as "From now on, always attach the SOC 2 report to an intake request", and every later run reads it. ## Memory Open **Configuration > Coverbase AI Memory**. Each row is one preference, with where it came from and how many runs used it. ### Where preferences come from | Source | When it is added | | - | - | | From a risk review | You moved an inherent risk level and said why | | From a declined checkpoint | You declined a step the run asked about | | From a change you kept as it was | You kept a record as it was instead of applying a proposed change | | From an instruction to a run | Your instruction started with "always", "never" or "from now on" | | Added here | You typed it into **Tell Coverbase AI something to remember** | | From the guided setup | It came from your organization's setup | To add one by hand, type it into **Tell Coverbase AI something to remember** and click **Add to memory**. ### Choose where a preference applies Set **Applies To** on a preference to limit it to one kind of run, such as **Onboarding a vendor**, **Reviewing assessments** or **Radar monitoring**. **Every run** applies it everywhere. ### Edit, turn off or forget * **Edit** changes the wording. * Turn off **Use this preference** to stop runs reading it while keeping it for later. * **Delete** forgets it. No run reads it again, though a later correction can teach it again. A run's report lists the preferences that changed what it did under **Because you told me**, with a link back here. ## What Coverbase AI may do on its own Under **What Coverbase AI May Do on Its Own** on the same page, each kind of change has a policy. It applies when a run is set to **Don't stop to ask**: | Policy | What happens | | - | - | | Applies on its own | The run makes the change without asking | | Always asks | The run stops and asks, even when told not to | | Off | The run never makes this change | Archiving a record, submitting an intake request and anything else that cannot be undone always ask, and their policy cannot be changed. ## Standing agents A standing agent keeps working without being asked: on a schedule, or when something happens in Coverbase. Open **Configuration > Coverbase AI Agents**. ### Add an agent Click **New agent**, or click **Add** on a template under **Templates**. A new agent starts turned off. | Field | What it sets | | - | - | | Name | What the agent is called in lists and notifications | | Standing Instructions | The job, in your own words, such as "Clear the Radar alert queue and ask me only about the alerts that change a vendor's residual risk" | | Owner | The person or group it acts as, with their permissions. A group's agent asks whoever is on duty under the group's assignment rules. Leave it empty to own it yourself. | | Escalate After (Minutes) | When nobody answers in time, the question moves to the next person on duty | | Schedule | Every day, every weekday, every Monday, the 1st of every month, or a custom schedule, at a time and timezone you choose | | Starts When | Events that start a run, each with the record it is about | | Playbooks | The playbooks it may use, or any playbook | | Approval | **Ask before writes**, or **Don't stop to ask** | | Runs per Day and Spend Cap per Day | Daily limits. Caps reset at midnight in the agent's timezone. | | Quiet From and Quiet Until | Hours when it starts no scheduled run. Anything it would tell you waits until they end. | **Starts When** can be: * A Radar alert comes in * A finding goes overdue * A contract's notice date comes due * An intake request is submitted * An assessment finishes Click **Create agent**, check the settings, then turn it on. ### Manage agents The list shows each agent's runs and spend today. From an agent's actions menu you can **Edit**, **Pause**, **Resume** or **Delete** it. Deleting an agent stops it starting runs, and the runs it already started stay with the people they ran for. **View in Agent Ledger** shows everything it did. When a scheduled run did not start, the row says why, such as "it fell in quiet hours" or "it had reached its spend cap for the day". **Pause every agent** stops all of them at once. Nothing starts, and live runs stop before their next step, until an admin turns it off. ## Troubleshooting | Problem | What to do | | - | - | | A run ignored a preference | Check that it is turned on and that **Applies To** covers that kind of run. | | A run keeps doing something you corrected | Add it as a preference that starts with "always" or "never", or set the change's policy to **Always asks**. | | An agent never runs | Check that it is turned on, that it has a schedule or an event under **Starts When**, and that every agent is not paused. | | An agent stopped partway through the day | It reached **Runs per Day** or **Spend Cap per Day**. The row says which. | | Nobody answered an agent's question | Set **Escalate After** so it moves to the next person on duty. | ## Related > **[Working with Coverbase AI](https://docs.coverbase.com/user-guides/coverbase-ai.md)** > > Start a run, answer what it asks, steer it and read the report. > **[Files and records](https://docs.coverbase.com/user-guides/coverbase-ai-files-and-records.md)** > > Attach documents and point a run at the right records. > **[Agent Ledger](https://docs.coverbase.com/user-guides/agent-ledger.md)** > > Every action an agent took, and who decided. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.