Integrations

Use Hackaiz from Claude, Cursor and your assistants

Connect your Hackaiz account to an MCP-compatible assistant to read your platforms, verified sites, analyses and findings. Reading is free; a component check runs only after you approve it in the conversation and costs credits based on the real work. The assistant never writes to GitHub.

Updated: November 21, 2026A service operated by HACKTUALIZ Inc.

MCP address: https://hackaiz.com/mcp. You sign in with your Hackaiz account (OAuth) and approve a consent screen. No key to copy.

  • Open Settings → Connectors → Add custom connector.
  • Paste https://hackaiz.com/mcp, then sign in to Hackaiz and approve access.
  • Ask for example: "List my verified Hackaiz sites" or "Explain the most severe finding of my latest analysis".
  • Claude Code: claude mcp add --transport http hackaiz https://hackaiz.com/mcp
  • Cursor / Windsurf / VS Code: add an HTTP MCP server pointing to https://hackaiz.com/mcp in the editor's MCP configuration.
  • On first use, the editor opens the Hackaiz sign-in in your browser.
  • List your platforms and verified sites.
  • Read an analysis status and its findings (no raw evidence or secrets).
  • Explain a finding.
  • Run an analysis (run_analysis) after you approve it in the assistant, on a target already proven or connected; the actual credit total is shown afterwards. Or prepare the Hackaiz link to launch it yourself.
  • check_component: checks a Skill, MCP server, package or workflow you paste, after you approve it in the assistant. Spends credits; static analysis only, nothing is executed.
  • Launch an analysis or spend credits without your explicit approval.
  • Analyse a target whose ownership you have not proven.
  • Write, merge or change anything on GitHub.
  • Take payments: purchases happen only on Hackaiz.

Assistants that accept a remote MCP server can use the same address, subject to their own rules and availability. ChatGPT apps and listing in official directories depend on each platform's approval and are not available yet.

Remove the connector in your assistant; access then stops for that assistant. Email contact@hackaiz.com with any question.

  • In your repository: curl -fsSL https://hackaiz.com/hackaiz-pre-commit.sh -o .git/hooks/pre-commit && chmod +x .git/hooks/pre-commit
  • The check runs on your machine and sends nothing. It catches common key formats (GitHub, AWS, Stripe, Supabase, private keys); it does not guarantee a commit is safe.
  • Download https://hackaiz.com/hackaiz-github-action.yml and place it at .github/workflows/hackaiz.yml.
  • Read-only: it looks for added secrets and applies the open Hackaiz rules, never writing or merging. A passing run does not prove the code is safe.
  • curl -fsSL https://hackaiz.com/hackaiz-scan.mjs -o hackaiz-scan.mjs && node hackaiz-scan.mjs .
  • Looks for secret keys in your files without sending anything; values are masked. No result does not prove the project is safe.
  • These tools accept remote MCP servers in their MCP settings: add an HTTP server named "hackaiz" with the address https://hackaiz.com/mcp.
  • The Hackaiz sign-in opens on first use; the same read tools and rules apply.
  • Exact support (OAuth, remote servers) depends on each tool's version; if yours lacks it, use Hackaiz from the dashboard.
  • In ChatGPT: Explore GPTs → Create → Configure → Create new action → Import from URL: https://hackaiz.com/openapi-gpt.json
  • Authentication: API key, "Bearer" type, with your platform key (platform page in Hackaiz).
  • Read-only: the GPT reads one platform's analyses and findings. Every access appears in Settings → Integrations.
  • Settings → Integrations: paste an incoming webhook address to receive serious proven risks.
  • On each finding: "Create Linear ticket" or "Copy for Jira". Nothing is sent without you.
  • CSV report download: importable into Jira or Linear (billed like other downloads).
  • JSON download: CycloneDX 1.5 inventory of npm, Python/PyPI, Go, Rust/Cargo, PHP/Composer and Ruby/RubyGems packages read from manifests and lockfiles during new repository analyses. Only packages actually read are included; other ecosystems and vendored code are not inventoried. Truncation remains visible: this is not a complete or certified SBOM.

The connected client is prepared under tools/hackaiz-cli; it is not published on npm. It submits an autonomous processing request and reads its status without opening Hackaiz. A personal key with read and launch scopes, an authorized platform and current proof and consent are required. Normal credits apply; purchases remain on Hackaiz.

Passive analysis only. The autonomous service must be published and synced before use. Closing the client does not request cancellation; started analyses are never automatically replayed after interruption. After a connection interruption, read the same request's status instead of creating a new request. No real customer-account test has been confirmed.

Template: https://hackaiz.com/hackaiz-connected-action.yml. It uses the prepared client, a personal key and your saved platform; it can comment a result link without modifying code or merging. Fork pull requests are excluded.

This template analyses the saved target, not necessarily the pull request commit. It is not verification of that commit. Real GitHub execution remains unverified.

The read_fix_preview tool reads the status and commit anchor of an existing prepared correction. Files, review and publication approval stay in Hackaiz. This tool does not prepare a new correction or publish anything.

The VS Code/Cursor extension (version 0.4.0) is published on Open VSX under the verified "Hackaiz" publisher: https://open-vsx.org/extension/Hackaiz/hackaiz-editor. Install it in Cursor and Windsurf by searching "Hackaiz" in Extensions.

It can store a personal key, read findings, request a passive analysis with editor confirmation and open approvals in Hackaiz. It does not modify files or write to GitHub. Published and synced autonomous processing is required.

  • Published on GitHub Marketplace: https://github.com/marketplace/actions/hackaiz-import-gate (repository Hacktualiz/import-gate).
  • Add the HACKAIZ_API_KEY secret (Settings → Secrets and variables → Actions), then a .github/workflows/hackaiz.yml file containing: uses: Hacktualiz/import-gate@v1 with api-key: ${{ secrets.HACKAIZ_API_KEY }}.
  • GitHub's "Choose a workflow" page does not list actions: you write the file yourself.
  • Danger or an incomplete result fails the job by default. Each run costs credits. Never expose the secret to fork pull requests.
  • The SARIF report can be sent to GitHub's Security tab; this depends on the repository's permissions and Code scanning settings.
  • Install: mkdir -p ~/.claude/skills/hackaiz-security && curl -fsSL https://hackaiz.com/skills/hackaiz-security/SKILL.md -o ~/.claude/skills/hackaiz-security/SKILL.md
  • Set HACKAIZ_API_KEY in your terminal, then restart Claude Code.
  • The Skill asks for your approval before every paid check and never sends .env files, keys or certificates.
  • In Claude web or Claude Desktop, a Skill cannot call Hackaiz: use the connector (section 2).