Docs/kernel getting started#14
Open
caffeinum wants to merge 31 commits into
Open
Conversation
eval showed agents using Claude OAuth tokens instead of GitHub PATs. added callouts, token prefix hints (ghp_), headless env note, and troubleshooting table. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
replaces Mintlify config with full Next.js + Fumadocs scaffold: - app router with docs layout, search API, and static generation - MDX content moved to content/docs/ with meta.json navigation - shim components for all Mintlify primitives (Card, Tabs, Steps, ParamField, etc.) - tailwind + fumadocs-ui theming with indigo primary color - internal links updated to /docs/ prefix, api: frontmatter stripped Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
simpler flow: gh auth login → gh auth token Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
covers: browser creation via CDP, playwright integration (TS + Python), CLI deployment, troubleshooting common eval errors Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
full getting-started docs: browsers, apps/CLI, troubleshooting, API reference. TS + Python examples throughout. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- quickstart: use playwright-core not playwright for deploys - browsers: add CDP retry logic, note about proxy readiness - apps: page.evaluate string literal gotcha, playwright-core warning - troubleshooting: add bundler bug, CDP race, deploy failures, correct doc URLs Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
agents guess /getting-started, /docs/browsers/sessions, /docs/sdk — redirect all to correct paths. llms.txt provides correct paths, inline quickstart, and key gotchas. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
shows top 3 most similar pages when agent hits a wrong URL. pure client-side, zero deps, instant. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
agents use WebFetch which throws on 404 status. middleware intercepts unknown paths and returns 200 with plain text suggestions (like vercel docs). keeps llms.txt link at bottom. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- redirects: /api-reference, /cloud-browsers, /browsers, /apps → /docs/* - api-reference: full parameter tables for Kernel() and browsers.create() with types, defaults, and return types - llms.txt: detailed descriptions per page for agent decision-making Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
the #1 agent friction point: timeout is silently ignored, must use timeout_seconds. also added inline CDP retry to quickstart example so agents don't have to discover it themselves. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
agent was making 8 fetches across 7 pages to assemble info. quickstart now has everything: install, auth, browser creation, full API reference with param tables, deployment with example app, and troubleshooting table. one fetch instead of seven. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- redirects: /getting-started/*, /reference/*, /guides/* all → correct docs paths
- quickstart: add app.action<{ url: string }>() typed payload example
- prevents the 3 URL 404s and TS compilation errors from latest eval
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
9.8KB comprehensive page caused agent to over-think (spawn 2 sub-agents, attempt deployment, handle edge cases). back to ~2.5KB focused quickstart: install → auth → create browser → done. typed action and full reference stay on api-reference page. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
quickstart is now one complete copy-paste file that actually runs (creates browser, scrapes HN, prints titles, cleans up). no patterns, no explanations between steps. also adds /.well-known/llms.txt → /llms.txt. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
the concise 4-step quickstart scored C+ (68, 17 calls). every subsequent change made it worse. reverting to the proven version while keeping redirects and llms.txt improvements. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
CDP connectOverCDP() caused 300s timeouts in 2/3 trials. browsers.playwright.execute() handles connection + cleanup automatically. also embedded 5 gotchas from llms.txt directly in quickstart (trial 3 scored highest because it read llms.txt — now all trials get it). added dashboard.onkernel.com domain banner. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
self-contained SKILL.md with setup, all examples, API reference, troubleshooting. one file replaces 4-8 WebFetch calls. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
|
Ready to review this PR? Stage has broken it down into 8 individual chapters for you: 
Chapters generated by Stage for commit 260468e on May 15, 2026 1:36am UTC. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Uh oh!
There was an error while loading. Please reload this page.