# FAQ for IBM Paver Find answers to frequently asked questions about using IBM® Paver, managing interactive sessions, working with credentials safely, configuring memory, and applying safety policies. > **Pro tip** > > Paver is the best guide to its own capabilities, so any questions that you have you can ask Paver in your interactive chat session in the web UI or by using `paver chat` in the CLI. ## Can I try Paver for free? Yes, you can try Paver at no cost today in [IBM Cloud®](https://paver.cloud.ibm.com/). You must have an IBM Cloud account to log in and access Paver. For more information about using Paver in IBM Cloud, go to the [IBM Cloud docs](https://cloud.ibm.com/docs/paver). ## What's the difference between `paver ask` and `paver chat`? `paver chat` starts a full interactive session with streamed responses, rendered markdown, and supports slash commands. `paver ask` sends a single message, streams the response, and exits. It's designed for scripts and automation, and supports `--json` output. For more information, see the [IBM Paver CLI command reference](https://paver.ibm.com/docs/cli-reference.md). ## What stops Paver from breaking something? Three layers of protection work together to provide safe guards: - **Safety mode** controls whether Paver can make changes in your environment. In Safe mode, Paver can't make changes, but it can still read your environment and read and write files in the sandbox. In Edit mode, Paver can make changes with your approval before it runs. - **Paver Guard** applies a tunable risk policy on top of Edit mode, such as asking Paver to try another approach before it attempts a high-risk operation. - **Access controls in the host offering** are the ultimate safeguard. Paver acts with your permissions, so it can't do anything that you can't do. For more information, see [Managing safeguards and controls in Paver](https://paver.ibm.com/docs/core-concepts/managing-risk-and-control.md). ## How do I give Paver credentials safely? Use the keyring. Store a secret once from an environment variable, a file, or an interactive prompt, and then export it. Paver then injects it into the sandbox as an environment variable. Values are encrypted, never shown after storage, and automatically redacted from Paver's output if they are 8 characters or longer. Never paste credentials directly into the chat. For more information, see [Storing your secrets in the keyring](https://paver.ibm.com/docs/core-concepts/keyring.md). ## Who can see my memories and files? Memories and sandbox files are scoped to your user account and isolated to your session, so other users can't see them. Only you and Paver have access to complete your tasks. For more information, see [Building persistent memory and context in Paver](https://paver.ibm.com/docs/core-concepts/memory.md) and [Sharing files with Paver](https://paver.ibm.com/docs/core-concepts/files.md). ## Can I work within Paver without using the agent? Yes, you can work with Paver without the agent in two different ways: - Prefix input with `!` or `bash` in chat to run a single command directly in the sandbox, bypassing the agent. - Open a full interactive terminal in the sandbox: - Use `/shell` from the web UI or an interactive chat in the CLI. - In the CLI, use `paver shell`. ## Does Paver remember what I told it last week? Yes. Paver automatically extracts facts worth remembering from your conversations, such as preferred regions, active projects, and workflow preferences, and stores them as memories to use in future sessions. It deliberately won't store sensitive values like API keys or passwords; use the Keyring for those. For more information, see [Building persistent memory and context in Paver](https://paver.ibm.com/docs/core-concepts/memory.md). ## How do I view, correct, or delete what Paver remembers? You can manage your memories from a chat session or from the CLI: - View memories with `/memory list` in a chat session or `paver memory list` in the CLI. - Correct a memory by telling Paver in the chat, for example, "Actually, my preferred region is eu-gb, not us-south." Paver updates the memory for you, so you don't need to delete the old one. - Add new memories with `/memory add ` or `paver memory add ` to save a memory directly. - Delete memories with `/memory remove ` to delete one memory or `/memory clear` to delete them all. In the CLI, use `paver memory remove ` or `paver memory clear`. The CLI asks you to confirm unless you add `-y`. Clearing your memories doesn't affect your conversation history, which Paver stores separately. You can also back up your memories by exporting them, and then restore them later by importing them. For more information, see [Building persistent memory and context in Paver](https://paver.ibm.com/docs/core-concepts/memory.md) and the [IBM Paver CLI command reference](https://paver.ibm.com/docs/cli-reference.md). ## How can I add my team's procedures or specialized knowledge? Package the knowledge as a skill. A skill is a folder that contains a `SKILL.md` file with instructions and optionally, scripts, references, and templates. Skills work well for workflows or internal tools that Paver doesn't know about by default. When your request matches a skill's description, Paver loads the skill automatically, so you don't need to select it. For more information, see [Extending Paver with skills](https://paver.ibm.com/docs/core-concepts/skills.md). ## Where is my data stored and how long is it retained? Paver stores your conversations, memories, skills, and files. The data is scoped to your user account. Paver encrypts data in transit with TLS and encrypts keyring secrets at rest. Paver keeps your conversations, memories, files, and skills for 90 days from the last time you access them. If you don't access your data within 90 days, Paver deletes it automatically. You can delete your data sooner. For example, run `paver conversation remove ` or `paver conversation clear` to remove conversations, and run `paver files remove ` to remove files. If you need to keep a record for longer than 90 days, export your conversations and store them in your own record-keeping system before the 90 days pass. For more information, see [Logging, audit, and data retention in Paver](https://paver.ibm.com/docs/security/logging-and-data.md) and [Understanding IBM Paver security and data protection](https://paver.ibm.com/docs/security/overview.md). ## Is my data used for model training? No. Paver never uses your prompts, context, or operational data to train foundation models. ## How do I report a problem? Run `/feedback` from the interactive chat session in the web UI: ```text /feedback create "your feedback here" ``` Or from the CLI: ```text paver feedback create "your feedback here" ``` Your feedback is submitted automatically along with the conversation context.