Add Computer Use overview page

[AnnaXWang]

Summary

Adds a landing/overview page for the Computer Use integrations group, which previously had six model pages (Anthropic, Gemini, OpenAGI, OpenAI, Tzafon, Yutori) but no entry point.

The overview page:

• Explains what computer use models are and the shared screenshot → predict → execute → repeat action-observation loop.
• Notes that the loop runs on Kernel's Computer Controls API (OS-level input, not CDP) and how that ties into bot detection.
• Links each supported model in a CardGroup, and notes any VLM works via Computer Controls.
• Includes a quick kernel create get-started, the standard Kernel benefits block, and next steps.

Build your own agent (cua-agent)

• Adds a "Build your own agent" section to the overview documenting the @onkernel/cua-agent library path: the npm install, a starter snippet, and a provider/model-ref table (Anthropic, OpenAI, Gemini, Tzafon, Yutori).
• Each cua-supported provider page now shows a provider-specific CuaAgent snippet directly below its template (with that provider's model ref). The npm install lives once in the overview, which each snippet links to. OpenAGI is intentionally left template-only since cua-agent doesn't support it.
• Note: @onkernel/cua-agent is TypeScript-only (no Python package), so these snippets are TS-only by necessity; the overview labels it a TypeScript library. The kernel create templates remain bilingual.

Wiring

• Added integrations/computer-use/overview as the first page of the Computer Use group in docs.json (matches the convention used by integrations/vercel/overview, auth/overview, etc.).
• Added a redirect from /integrations/computer-use → /integrations/computer-use/overview so the bare section URL resolves to the new page.

Notes / test plan

• docs.json validated as well-formed JSON.
• Internal links verified against existing pages; provider snippets link to the overview's #build-your-own-agent anchor.
• cua-agent snippets, model refs, and env vars verified against kernel/cua (packages/agent README and packages/ai model catalog).
• The Mintlify CLI isn't available in this environment, so I couldn't run mintlify dev for a live preview. Recommend a quick visual check via the branch preview before merging.

Preview: https://tbd-6fc993ce-hypeship-computer-use-overview.mintlify.app/integrations/computer-use/overview

---

Note

Low Risk
Documentation and navigation-only changes with no runtime, auth, or data-handling impact.

Overview
Adds a Computer Use landing page and wires it into the docs nav so /integrations/computer-use resolves to the new overview.

The overview explains the shared screenshot → predict → execute loop on Kernel’s Computer Controls API, links all six model integrations, and documents the @onkernel/cua-agent TypeScript path (npm install, starter code, provider model refs, and env vars).

Anthropic, Gemini, OpenAI, Tzafon, and Yutori pages each gain a Build your own agent section with a provider-specific CuaAgent snippet pointing back to the overview anchor; OpenAGI stays template-only. Intro copy on a few pages drops em dashes for plain “by” wording.

^{Reviewed by Cursor Bugbot for commit 35bf54b. Bugbot is set up for automated code reviews on this repo. Configure here.}

[mintlify]

Preview deployment for your docs. Learn more about Mintlify Previews.

Project	Status	Preview	Updated (UTC)
Kernel	🟢 Ready	View Preview	Jun 29, 2026, 11:20 PM

💡 Tip: Enable Workflows to automatically generate PRs for you.

[cursor]

Stale comment

Risk assessment: Very Low

I reviewed the actual diff for the current head (3780ea9f5a0fadf3a7f00990dcba6fa28df4d7da). The PR adds a new integrations/computer-use/overview.mdx documentation page and wires it into docs.json with a navigation entry plus a redirect. Since the previous automation approval, the only additional change is a one-line docs link update from /browsers/computer-controls to /browsers/computer-controls#take-screenshots.

There are no production/runtime codepaths, infrastructure files, auth or permission logic, schema changes, shared services, or prompt/instruction files modified. I also found no CODEOWNERS file requiring owner review. Risk remains Very Low; the existing approval does not need to be revoked, and I am not re-approving because this PR is already approved.

  

_{Sent by Cursor Automation: Assign PR reviewers}

[cursor]

Stale comment

Risk assessment: Very Low

I reviewed the actual diff for the current head (4f00c550ecba52140ef87e6b7b761b7b807dd533). This PR adds a new integrations/computer-use/overview.mdx documentation page and wires it into docs.json with one redirect plus one sidebar entry. The latest synchronization since the prior automation assessment only changes punctuation and wording in that new docs page.

I found no production/runtime codepaths, infrastructure files, auth or permission logic, schema changes, shared services, or prompt/instruction files modified. I also found no CODEOWNERS file requiring owner review and no .git-blame-ignore-revs file affecting the diff. The PR is already approved, so I’m not re-approving it. Risk has not increased, so there’s no approval to revoke.

  

_{Sent by Cursor Automation: Assign PR reviewers}

[cursor]

Stale comment

Risk assessment: Very Low

I reviewed the actual diff for the current head (3393a5e05507dc46d94974d696b131fed8b4cffc). The PR changes docs.json navigation/redirects, adds integrations/computer-use/overview.mdx, and adds short cross-links from existing Computer Use integration docs to the new overview.

The latest synchronized change expands the new docs page with @onkernel/cua-agent install guidance, a TypeScript example, and provider model refs. That is still documentation-only: no production/runtime codepaths, infrastructure, auth or permission logic, schemas, shared services, or runtime prompt/instruction files are modified. I found no CODEOWNERS file requiring owner review and no .git-blame-ignore-revs file affecting the diff.

Risk remains Very Low. The PR is already approved, so I’m not re-approving it; because the risk did not increase, I’m leaving the existing approval in place.

  

_{Sent by Cursor Automation: Assign PR reviewers}

[cursor]

Stale comment

Risk assessment: Very Low

I reviewed the actual diff for head 3393a5e05507dc46d94974d696b131fed8b4cffc. The PR adds one Mintlify docs page at integrations/computer-use/overview.mdx, adds a sidebar entry and redirect in docs.json, and adds short cross-links from existing Computer Use integration docs to the new overview.

This is documentation/navigation-only: no production runtime codepaths, infrastructure, auth or permission logic, schemas, shared services, or runtime prompt/instruction files are modified. I found no CODEOWNERS file requiring owner review and no .git-blame-ignore-revs file affecting the diff.

Risk remains Very Low. The PR is already approved, so I’m not re-approving it; because the risk did not increase, I’m leaving the existing approval in place.

  

_{Sent by Cursor Automation: Assign PR reviewers}

[cursor]

Risk assessment: Very Low

I reviewed the actual diff for the current head fe1b639c1be14dce57462c1962f0d7e9025905b8. The PR changes Mintlify documentation/navigation only: it adds integrations/computer-use/overview.mdx, adds a redirect and sidebar entry in docs.json, and adds short cross-links from existing Computer Use integration docs to the new overview.

Since the previous assessment at 3393a5e05507dc46d94974d696b131fed8b4cffc, the only code diff is a two-link documentation adjustment in integrations/computer-use/overview.mdx that moves the screenshot anchor to the action-observation loop step. This does not alter runtime codepaths, infrastructure, auth or permission logic, schemas, shared services, or runtime prompt/instruction files.

I found no CODEOWNERS file requiring owner review and no .git-blame-ignore-revs file affecting the diff. Risk remains Very Low. The PR is already approved, so I’m not re-approving it; because the risk did not increase, I’m leaving the existing approval in place.

  

_{Sent by Cursor Automation: Assign PR reviewers}

[cursor]

Cursor Bugbot has reviewed your changes using high effort and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

Want fixes drafted automatically? Bugbot Autofix can create code changes for findings. A team admin can enable Autofix in the Cursor dashboard.

^{Reviewed by Cursor Bugbot for commit fe1b639. Configure here.}

[cursor]

Profiles linked instead of Managed Auth

Low Severity

The new Session state benefit points readers to Profiles for maintaining browser state across runs. Integration benefits that cover session persistence or authenticated browsing are expected to highlight Managed Auth (e.g. /auth/overview) as the primary path, not Profiles alone.

 

^{Triggered by learned rule: Integration pages should highlight Managed Auth for authenticated browsing}

^{Reviewed by Cursor Bugbot for commit fe1b639. Configure here.}

[dprevoznik]

This works well, but I could have been more clear in my suggestion. I think it's worth presenting the template installation at the top of each provider page and then for each individual provider page, we provide the provider specific snippet shown here: https://tbd-6fc993ce-hypeship-computer-use-overview.mintlify.app/integrations/computer-use/overview#build-your-own-agent (not the npm installs part just the TS)

Thoughts @AnnaXWang ? Happy to make the update if you agree / want me to

[AnnaXWang]

added in new commit!

[dprevoznik]

Same comment as Anthropic

[dprevoznik]

same comment as anthropic

[dprevoznik]

LGTM - left a suggestion

[cursor]

Risk assessment: Very Low.

Evidence from the exact PR diff (1baeb1e...3cad613): the changes add a new Computer Use overview docs page, add docs-only @onkernel/cua-agent snippets to existing Computer Use integration pages, and update docs.json with one redirect plus one navigation entry. I found no runtime code, production logic, infrastructure, auth, billing, permissions, schema, or shared library changes in scope.

Approval note: the PR is already in an approved review state, so I’m not adding another approval.

  

_{Sent by Cursor Automation: Assign PR reviewers}

[cursor]

Risk assessment: Very Low.

Evidence from the actual current diff (origin/main...origin/hypeship/computer-use-overview, head 35bf54b7fde9519c11e8847fbdcc1cc8c11f7e29): this PR modifies Mintlify docs/navigation only. It adds integrations/computer-use/overview.mdx, adds @onkernel/cua-agent documentation snippets to existing Computer Use integration docs, and updates docs.json with one redirect plus one sidebar entry.

I found no runtime codepaths, production logic, infrastructure files, auth, billing, permissions, database schema, shared libraries, or runtime prompt/instruction files modified. There is no CODEOWNERS file in the checked-out repository and no .git-blame-ignore-revs file affecting the assessment.

Approval note: the PR is already approved, so I’m not adding another approval. The latest synchronization does not increase risk, so there is no approval to revoke.

  

_{Sent by Cursor Automation: Assign PR reviewers}