L4
hardcoded_secretsrc/client.pyMove the credential to an environment variable before completing this change.
MCP privacy preflight
Give coding agents a privacy check they can call.
agentleak mcp exposes four local tools for code scanning, trace checks, redaction and change-aware preflight. Before an agent says the work is done, it can see what is new, what it fixed and what remains open.
$ privacy_preflight workspace1 new2 fixed1 still open
Four tools, one local engine
The MCP surface runs the same analyzer as the CLI, directly inside Claude Code, Cursor or any compatible client.
- privacy_preflight
- privacy_scan_code
- privacy_check_trace
- privacy_redact
Deltas an agent can act on
Preflight compares the current scan with the previous check and separates new, fixed and still-open findings instead of returning an isolated score.
- Stable identity across line moves
- Priority-sorted new findings
- Fixed findings acknowledged
- Empty traces rejected, never scored 100
Private by default
History stays in the project as capped, redacted JSONL. Without a project key, no result leaves the machine.
- No account required
- No network required
- Only redacted snippets in history
- Optional workspace sync
The mental model
A coding agent needs change context, not another dashboard.
A static score does not tell an agent whether its latest edit improved the system. Preflight makes privacy part of the completion loop: scan, identify the new disclosure, change the code, and verify that the same finding is gone.
Before
The agent finishes when the code compiles.
With AgentLeak
The agent finishes when its change introduces no new privacy finding.
- Finding identity survives harmless line-number changes
- Local history stores redacted evidence only
- Detection mode and tiers remain visible so a clean result is never overstated
How it works
From editor setup to a verified change.
1
Install the MCP extra
Install agentleak[mcp] in the same local environment your coding client can launch.
2
Register the server
Add agentleak mcp to the client MCP configuration. It communicates over standard input and output.
3
Run preflight
The agent scans the workspace and receives new, fixed and still-open findings with concrete remediation.
4
Fix and verify
Re-run the same tool before completion. A project key can optionally sync the result to a shared workspace.
Register the local MCP server
{
"mcpServers": {
"agentleak": {
"command": "agentleak",
"args": ["mcp"]
}
}
}FAQ
Questions, answered.
Ready when you are
Test the path, not only the answer.
Create a local workspace, run a bundled scenario, then wire AgentLeak into CI or let your agent onboard itself.