> ## Documentation Index > Fetch the complete documentation index at: https://docs.peepsai.com/llms.txt > Use this file to discover all available pages before exploring further. # Permissions and approvals > What each permission on the consent screen allows, which Peeps role you need, and what your agent asks you before it changes anything. Three things decide what your agent can do in Peeps: the permissions you approved when you connected, your role in the organization, and your approval in the conversation. ## The five permissions The consent screen lists these, in this wording. | Permission | What the consent screen says | What it unlocks | | ------------------ | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------- | | **Read** | Read your test cases, runs and results | Every read: failures, evidence, test cases, scripts, folders, environments, projects. | | **Run** | Run your tests, including against production environments | Running a test, verifying a fix, and converting a script's language. | | **Edit** | Create and edit test cases, folders, cleanup links, environments, variables, projects and test scripts, archive and restore test cases, archive script versions, delete empty folders, submit fixes and stage imports | Every change that is not a publish. | | **Publish** | Generate and publish test scripts, publish test scripts you approved, publish imported scripts you approved, and archive the version that runs, or restore any archived version | Anything that changes which script version your runs use. | | **Stay connected** | Keep this connection without re-approving | Your client renews its access without sending you back to the browser. | Some actions need more than one. Submitting a fix needs **Edit** and **Run**. Publishing a verified fix needs **Edit** and **Publish**. Generating a script needs **Edit**, **Run** and **Publish**. The [command reference](/commands) lists what each command needs. To see what your connection carries right now, ask your agent *"who am I in Peeps?"*. A permission covers a kind of action, not a fixed list of commands. If Peeps adds a new editing command later, a connection that already holds **Edit** can use it without asking you again. If your agent reports that the connection lacks a permission, reconnect from Peeps settings, under **Coding agents**. Peeps cannot widen an existing connection from its side. ## Your role still applies Permissions never lift you above your role in the organization. | Role | What your agent can do for you | | ----------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------- | | **Member** | Read everything, and run tests. | | **Contributor** and **Owner** | Everything above, plus every change: creating and editing tests, fixing and publishing, environments, folders, projects and Linear links. | ## What your agent asks you first Your agent shows you what it is about to do and waits for your approval before it changes anything. Approving one step never approves the next. Running a fix, publishing it, and archiving something are separate decisions. For the actions with the most reach, Peeps ties your approval to exactly what you saw: * **Runs, fix verifications and script generations** spend a plan you approved. The plan names the environment, says whether it is production, and lists every script that will run. A plan is single-use and expires after 30 minutes. If the environment, its variables or the tests around it change before it is used, Peeps refuses it, and your agent asks you to approve a new one. * **Publishing** binds the exact script version you reviewed and the version it replaces. If either one moved since you looked, Peeps refuses and your agent shows you the change again. * **Edits to steps, details and prerequisites** carry a fingerprint of what your agent read. If somebody else edited the test case in the meantime, Peeps refuses rather than overwriting their work. For other changes, such as creating a test case or a folder, your agent asks you before it acts. Peeps records the change as yours. It cannot see your conversation, so it cannot check that you said yes. Read what your agent proposes before you approve it. ## Production Peeps treats an environment named `production` or `prod` as production. Your agent tells you when a plan targets production, and a change to a production environment needs your explicit confirmation. ## Who else hears about it Your organization's owners and admins are notified when a coding agent: * publishes a script version it added, or a verified fix * creates an environment, or sets, changes or deletes one of its variables * renames an environment or changes its details ## Where the MCP server stops * **Web first.** Running, fixing, generating, importing and editing scripts work on web (Playwright) test cases. Archiving, restoring and editing a test case's details also work for iOS and Android test cases. * **No project deletion.** No coding agent can delete a project. * **Projects whose tests live in your repository.** If your project runs specs from your own GitHub repository in your own CI, change those specs in your repository. Peeps refuses to create test cases, or to run, fix, generate, import or publish scripts, for that project over MCP. Archiving and restoring test cases still works. It changes how Peeps organizes its test cases, never the repository. Editing a test case's steps also works, but it edits the Peeps-side notes only. * **Some settings stay in the web app.** A project's timezone, auto-retry and auto-heal settings, file variables, and mobile projects are managed in the Peeps web app. So is connecting Linear. ## What leaves your machine * **A coverage check** sends search terms only. * **Creating a test case** sends the steps you approved plus a commit reference. * **Adding or importing a script** sends the test files you approved. Peeps screens each one for credentials. * **Nothing else.** There is no upload of your checkout in any flow. If your organization has chosen to keep the source of tests it turns into plain-English cases, your agent also sends those original files when you ask it to convert them. Peeps keeps them and does not show them back. In the other direction, failure evidence — including screenshots, traces, videos, and DOM snapshots of your application — reaches whichever AI provider your client is configured with, under that provider's own terms. ## Revoking a connection Peeps settings, under **Coding agents**, lists every client you have authorized. Revoking one stops it renewing its access. A token already in that client's hands stays valid until it expires, up to 15 minutes later. Revoking applies to every installation authorized under that client, on every machine.