Install anywhere
you ship code.
Pick your stack. Each install takes under 30 seconds. Same token works across every channel.
0. Get your token
Start here- Sign in at literalsec.com/install (Google or GitHub).
- Copy your project token from the install page.
- Save it once — every channel reads
~/.literalsec/token.
1. CLI · the universal installer
RecommendedOne command. Auto-detects git, VS Code, Claude Code, Cursor in your project — wires each.
curl -fsSL https://literalsec.com/install.sh | bash
Then, in any project:
literalsec init
Other commands:
literalsec login— save your token.literalsec status— what's wired here.literalsec scan path/to/file.ts— manual scan.
2. MCP server · for AI agents
CoreAdds a secure tool the AI can call. Works in Claude Code, Cursor agent mode, Windsurf, Cline, Continue, Roo Code, Goose. Also enables /secure.
Claude Code
claude mcp add --transport http --scope user literalsec \ https://literalsec.com/mcp \ --header "Authorization: Bearer YOUR_TOKEN"
Cursor / Windsurf
Add to .cursor/mcp.json or .windsurf/mcp.json:
{
"mcpServers": {
"literalsec": {
"url": "https://literalsec.com/mcp",
"headers": { "Authorization": "Bearer YOUR_TOKEN" }
}
}
}
Cline · Continue · Roo · Goose
Same MCP shape. URL https://literalsec.com/mcp + Authorization header.
3. VS Code extension · universal save-time scan
UniversalThe extension catches both editor saves (Cmd-S) and direct file writes (AI tools that bypass the editor). Scoped to the open workspace — never scans outside your project root. Findings appear in the Problems panel.
Manual install — download the .vsix
The extension isn't published to the Marketplace yet (coming soon). Install manually:
- Download literalsec-1.0.6.vsix (v1.0.6)
SHA-256 in SHA256SUMS · verify withsha256sum literalsec-1.0.6.vsix - One-liner from terminal (works for VS Code / Cursor / Windsurf / Codespaces / VSCodium):
code --install-extension literalsec-1.0.6.vsix
Cursor:cursor --install-extension literalsec-1.0.6.vsix· Windsurf:windsurf --install-extension literalsec-1.0.6.vsix - Reload the editor when prompted
- Open Settings → Extensions → Literal Security and paste your token (or save it once to
~/.literalsec/token)
Cursor (GUI)
- Cmd-Shift-P (or Ctrl-Shift-P) → "Extensions: Install from VSIX…"
- Pick
literalsec-1.0.6.vsix - Reload window when prompted
- Settings → Extensions → Literal Security → Token
Windsurf (GUI)
- Cmd-Shift-X → click the … menu (top-right) → "Install from VSIX…"
- Pick
literalsec-1.0.6.vsix - Reload
- Settings → Extensions → Literal Security → Token
VS Code · GitHub Codespaces · VSCodium · Gitpod
- Cmd-Shift-X → … menu → "Install from VSIX…"
- Pick
literalsec-1.0.6.vsix - Reload
- Settings → Extensions → Literal Security → Token
Cline · Continue · Roo Code · GitHub Copilot · Codeium
These tools run as VS Code extensions themselves. Install our extension in the same VS Code (or Cursor / Windsurf) instance — both run side-by-side. When the agent writes a file via its internal write tool, our FileSystemWatcher catches it.
Zed · Claude Code (CLI) · Aider
Zed doesn't support VS Code extensions natively; Claude Code CLI and Aider are CLI tools. Use the MCP server for Claude Code, our CLI for Aider, and either MCP or the git hook for Zed.
4. JetBrains plugin · IDE family
BetaIntelliJ IDEA, WebStorm, PyCharm, GoLand, RustRover, RubyMine, PhpStorm, CLion, Android Studio.
Manual install — download the .zip
- Download literalsec-jetbrains-1.0.2.zip (v1.0.2)
SHA-256 in SHA256SUMS · verify withsha256sum literalsec-jetbrains-1.0.2.zip - Open any JetBrains IDE → Settings → Plugins → click the ⚙ at the top → Install Plugin from Disk…
- Pick
literalsec-jetbrains-1.0.2.zip - Restart the IDE when prompted
- Paste your token in Settings → Tools → Literal Security, or reuse
~/.literalsec/token
5. Git pre-commit hook · universal backstop
UniversalWorks for any AI tool, any editor, any workflow — as long as you commit via git. Blocks commits with critical findings.
curl -fsSL https://literalsec.com/install-git-hook.sh | bash
Run from any git repo. Bypass per-commit: LITERALSEC_SKIP=1 git commit ...
6. The /secure command
Auto-included
After installing the MCP server, type /secure in any AI session. The AI scans every recently-edited source file and surfaces findings with title, file:line, evidence, and a concrete fix.
~100% invocation across every major AI coding assistant — frontier and open-source.
7. Add your site for probes
Paid tiersProbes hit your deployed site for runtime issues. Findings route back to your IDE so the AI fixes the source.
- Dashboard → set your project's Site URL.
- (Optional, Startup+) Add authenticated session creds for behind-auth routes.
- Run probe or wait for the weekly / daily auto-scan.
Quotas: Solo 1/mo · Startup 2/mo · Business unlimited.
What we catch
- Hardcoded secrets (Stripe, AWS, GCP, OpenAI, Anthropic, GitHub — 700+ patterns)
- SQL / NoSQL / shell injection
- IDOR + broken auth (route lacks ownership check)
- XSS sinks, SSRF, open redirects, CORS-with-credentials
- Weak crypto (MD5/SHA1 for passwords,
Math.randomfor tokens) - Deserialization, path traversal, prototype pollution
- Typosquatted deps, known CVEs, RLS gaps
- Chained attacks — e.g. "SQL injection + missing auth = unauthenticated DB exfil"
FAQ
Do you train AI on my code?
No. Source code is stateless and never persisted. Only the audit log of findings is stored.
What if I'm offline?
Clients fail-open. Your save / commit proceeds. The next online scan picks up.
Can I bypass for a single commit?
LITERALSEC_SKIP=1 git commit ...
How do I rotate my token?
Dashboard → Rotate token. Update ~/.literalsec/token and any MCP configs that hardcoded it.