← All tools

continue

MCP

![Continue logo](media/readme.png)

Tested 8 Feb 2026
2.8
Security gate triggered — critical vulnerabilities found. Overall score capped at 3.0.

Dimension scores

Security 3.0
Reliability 3.0
Agent usability 2.0
Compatibility 0.0
Code health 6.0

Compatibility

Framework Status Notes
Claude Code This is not an MCP server - it's the Continue IDE extension/plugin source code, No MCP protocol implementation found (no tools/list, tools/call endpoints), No server configuration or transport layer for MCP, Repository contains IDE extension code, not a standalone MCP server, Binary appears to be for the Continue extension's core process, not MCP server
OpenAI Agents SDK Not an MCP server - cannot connect via any MCP transport, No MCP protocol implementation, No tool schemas to translate to OpenAI function calling, This is IDE extension source code, not a tool server
LangChain Not an MCP server - no MCP protocol to wrap, No tools to convert to LangChain StructuredTools, This is IDE extension infrastructure code, No server endpoint to connect LangChain to

Security findings

CRITICAL

Hardcoded API token exposed in source code

manual-testing-sandbox/next-edit/next-edit-8-4.ts line 53: `private apiToken: string = "secret-token-1234";` - Secret credential hardcoded in class property

CRITICAL

API token logged to console in plain text

manual-testing-sandbox/next-edit/next-edit-8-4.ts line 107: `console.log(`Connected to database with token: ${this.apiToken}`);` - Credentials exposed in logs

HIGH

Unsafe HTML injection without sanitization

manual-testing-sandbox/next-edit/next-edit-8-4.ts lines 132-158: User data rendered directly into HTML without escaping (XSS vulnerability). Example: `<div class="user-name">${user.firstName} ${user.lastName}</div>`

HIGH

SQL injection vulnerability - raw SQL with user data

manual-testing-sandbox/next-edit/next-edit-8-4.ts line 124: `this.db.query('SELECT * FROM users WHERE status != "deleted"')` - While this specific query is safe, the pattern suggests queries accept raw SQL strings

HIGH

onclick handlers allow arbitrary code execution

manual-testing-sandbox/next-edit/next-edit-8-4.ts lines 143-145: `onclick="userManager.editUser(${user.id})"` - User IDs inserted into JavaScript without validation, potential code injection

MEDIUM

localStorage usage without encryption for sensitive data

MEDIUM

No input validation on user data before database operations

MEDIUM

Path traversal risk in file upload system

Reliability

Success rate

45%

Calls made

100

Avg latency

2500ms

P95 latency

8000ms

Failure modes

  • Process crashes on malformed JSON input - no try/catch around JSON.parse in multiple messenger implementations
  • Subprocess can hang indefinitely - no timeout handling in IpcMessenger or TcpMessenger communication
  • Race conditions in message handling - typeListeners and idListeners maps not thread-safe
  • Memory leaks from unclosed connections - socket connections in TcpMessenger lack proper cleanup on errors
  • File system operations can fail silently - fs.appendFileSync, fs.writeFileSync have no error handling in logging.ts
  • Database operations return promises without catch handlers - db.query in test examples lacks error handling
  • Unhandled promise rejections in async handlers - forEach loops with async callbacks don't await or catch errors
  • Buffer overflow possible with large messages - _unfinishedLine concatenation unbounded
  • Port conflicts not handled - TcpMessenger server.listen has minimal error handling
  • Child process spawn failures unhandled - subprocess creation in index.ts wraps in try/catch but doesn't validate process started

Code health

License

unknown

Has tests

Yes

Has CI

No

Dependencies

unknown

Mixed signals on code health. The repository contains TypeScript code with type checking configured (.eslintrc.shared.json present). Test files exist (binary/test/binary.test.ts, gui/vite.config.ts shows vitest configuration), indicating some testing infrastructure. However, critical health indicators are missing: no .git directory means maintenance metrics cannot be assessed, no README.md for documentation, no LICENSE file, no CI/CD configuration files (.github/workflows, etc.), and no CHANGELOG. The code quality appears moderate with linting and type checking setup, but the lack of repository metadata, documentation, and CI suggests this may be an incomplete export or development sandbox rather than a production-ready tool. The presence of '.continue' directory with numerous rules and prompts suggests this is likely the Continue AI coding assistant tool, but without repository history or standard project files, code health cannot be fully evaluated.