Recipes
Short, single-task how-tos for peek. Each recipe walks you through one concrete workflow end-to-end.
Hero recipes
Let an AI agent assist a job application while you supervise
I want the agent to fill the boring parts of an application from my résumé — I review every step, I do the final submit myself, and I can stop it instantly.
Let Claude Code reproduce a bug on your authenticated staging dashboard
When my agent can't see my staging dashboard because it's behind SSO, I want it to look at the broken page I'm already on and fix it.
Clone a web app's UI from a real peek session
When I rebuild a clunky internal tool, I don't want my agent guessing at its structure from a screenshot. I want the real DOM the page actually rendered.
Generate a Playwright repro from a real browser session
When I just hit a bug by clicking around, I want an automated regression test before I forget how I did it.
Let Cursor see the real network calls your SPA is making
When my SPA's frontend is making twelve calls and one is failing, I want Cursor to read them and tell me which header is wrong.
Short recipes
Find a past session by what happened
When I vaguely remember a bug from a past session but can't recall the exact sessionId, I want my agent to search for it by keyword, origin, or date.
Verify and share a tamper-evident audit trail
When an agent has been acting in my browser, I want proof that the audit log is intact before I share it with a colleague or attach it to a bug report.
Set up peek with VS Code (Copilot) in 2 minutes
When I want peek's captured sessions in VS Code's Copilot agent, I want the one MCP entry that actually works — not a config VS Code silently ignores.
Using peek with your AI coding agent
Once peek is wired into my agent, I want to know when it'll actually help and which of its tools to ask for — without guessing.
Replay how the DOM changed around a failure
I want to see how the page actually changed over time — track one element's history, or every change in the seconds around a failure — from the recorded session.
Find the network call that broke your flow
When a page silently failed, I want my agent to find the request that returned an error — the method, URL, status, and what came back — without me digging through devtools.
Triage every console error in a recorded session
When something breaks in my browser, I want my agent to list every console error from the recorded session and tell me what I did right before each one.
See the causal chain that led to an error
I don't just want the error — I want the story: what I clicked, what the DOM did, and which request failed, in order, right before it broke.
Security-review a flow by letting your agent inspect the live DOM
When I'm reviewing a sign-up flow for tokens-in-URL or leaked secrets, I want my agent to look at the actual rendered DOM and network traffic — not the source.
Set up peek with Claude Code in 2 minutes
When I want to try peek with my Claude Code install, I want one command that wires up the MCP server without me touching JSON files.
Set up peek with Cline, Windsurf, or Codex CLI
When I use Cline, Windsurf, or Codex CLI, I want peek's MCP server registered without hunting for each tool's config file.
Set up peek with Cursor
When I'm a Cursor user, I want peek wired into my MCP config so the agent can read my browser sessions.
Understand peek's per-action approval model for sensitive flows
When I'm thinking about letting an agent drive clicks in a real browser, I want to know exactly what peek's permission model is — what it gates, what it logs, and what it never auto-approves.