Allows to propose removal pull requests for dead code (unused imports) on GitHub, after verifying the test suite passes.
Click on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@safe-fix-mcp scan repo for dead code and propose safe PR for unused imports"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
A real MCP server that finds dead code in a Python repository and proposes a branch + PR for the one class of finding that's actually safe to auto-remove: unused imports on a single-name import line. Everything else it finds — unused functions, classes, unused dependencies — stays report-only. A human always merges. This tool never merges anything itself.
Most "dead code" tools stop at reporting. The ones that go further usually delete with no safety net. This does neither: it drafts a real, minimal, reviewable PR, gated on the target repo's own full test suite actually passing after the removal — not a heuristic, a real pytest run. If the suite fails, nothing is committed or pushed; the repo is left exactly as it started.
Related MCP server: code-review-mcp-server
scan_dead_code(repo_path=".", min_confidence=60)Read-only. Runs vulture (dead code) and deptry (dependency issues) and returns a human-readable list of findings. Never modifies anything.
propose_removal_pr(repo_path=".")Refuses on a dirty working tree — never edits on top of uncommitted work.
Filters to unused imports at ≥90% vulture confidence, on single-name import lines only (from x import y, z is skipped — removing the whole line would silently remove z too).
Creates a real branch, removes the qualifying imports, runs the repo's real full test suite.
Only on a real pass: commits, pushes, and tries to open a PR via gh pr create.
If gh isn't installed/authenticated, the branch is still committed and pushed for real — only PR creation degrades, with the real error returned so you can open it manually.
pip install safe-fix-mcpAdd to your MCP client config (e.g. Claude Code):
claude mcp add safe-fix-mcp -- safe-fix-mcpOr run directly for local testing:
python -m safe_fix_mcp.serverInstalling from source instead of PyPI:
pip install -e .A SKILL.md ships inside this same package — not a separate download — and teaches Claude when and how to use the two tools correctly (e.g. that propose_removal_pr only ever auto-removes single-name unused imports, never functions/classes/dependencies). Install it into your project:
safe-fix-mcp-install-skillRun it from the project root you want the skill active in — it writes to .claude/skills/safe-fix-mcp/SKILL.md relative to your current directory.
Python ≥ 3.10
git on PATH
gh (GitHub CLI) on PATH and authenticated, if you want propose_removal_pr to actually open the PR — without it, the branch still gets pushed for real, and the tool tells you to open the PR by hand.
pip install -e ".[dev]"
pytestvulture flags scan_dead_code/propose_removal_pr themselves as "unused" — a known false-positive class, not a real bug: they're dispatched by the @mcp.tool() decorator at runtime, not called directly anywhere in the source, so static call-graph analysis can't see the real caller (the MCP framework itself).
scripts/verify_real_client.py launches the packaged server as a real subprocess and talks to it with the real mcp.client.ClientSession — the same path a real MCP client uses. Useful as a smoke test after any change:
python scripts/verify_real_client.pyPropose a real branch+PR removing only unused imports (single-name
import lines, >= 90% vulture confidence) from a Python repository. Runs
the repo's own full test suite as a safety gate before ever committing
or pushing — a failure there reverts everything and nothing is
committed or pushed. If gh isn't available or PR creation otherwise
fails, the branch is still committed and pushed for real; only the PR
itself needs opening manually. A human always merges — this never
merges anything itself.
repo_path: path to a real git repository with a clean working tree (uncommitted changes are refused, not overwritten).
| Name | Required | Description | Default |
|---|---|---|---|
| repo_path | No | . |
| Name | Required | Description |
|---|---|---|
| result | Yes |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations, the description fully carries the behavioral burden, and it does so well. It explains that the test suite is a safety gate, that failures revert all changes, that branch commits/pushes persist if only PR creation fails, and that human merge is always required.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is somewhat long but every sentence earns its place, covering prerequisites, side effects, failure modes, and parameter meaning. The critical safety guarantees are front-loaded, and there is no filler.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
For a tool with one optional parameter and an output schema, the description is remarkably complete. It covers prerequisites, refused inputs, failure behavior, push semantics, and merge policy, so an agent has enough to call it correctly without additional context.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema description coverage is 0%, but the description compensates by explaining repo_path as a real git repository with a clean working tree and noting uncommitted changes are refused. This adds meaningful semantics beyond the bare schema field, though it does not mention optionality/default behavior beyond what the schema already shows.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the tool's specific action: propose a real branch and PR that removes unused imports from a Python repository. It adds precise scoping (single-name import lines, >=90% vulture confidence) that makes its purpose concrete and distinct from a general cleanup tool.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description implies usage context by specifying criteria for eligible imports and requiring a clean git repository with a full test suite. However, it never explicitly addresses when to use this tool versus scan_dead_code or mentions exclusions, leaving the routing partially to inference.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Scan a Python repository for real dead code (vulture) and dependency issues (deptry). Read-only — never modifies anything. Returns a human-readable list of findings, or "No real candidates found."
repo_path: path to the repository to scan (must contain a pyproject.toml for the dependency checks to run; dead-code scanning works regardless). min_confidence: vulture's own confidence threshold (0-100). Lower values surface more candidates but more false positives.
| Name | Required | Description | Default |
|---|---|---|---|
| repo_path | No | . | |
| min_confidence | No |
| Name | Required | Description |
|---|---|---|
| result | Yes |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations present, the description carries full burden and does so well: it states the tool is read-only, never modifies anything, and describes the exact output (human-readable findings or 'No real candidates found.'). It also discloses the pyproject.toml dependency for deptry and explains the confidence threshold's trade-off. No annotation contradiction.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is compact—three sentences for behavior plus two parameter explanations—and all content earns its place, with read-only status front-loaded.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
For a two-parameter tool with an output schema and a single sibling, this description covers input semantics, prerequisites, behavioral safety, and result format. The only omitted piece is explicit sibling differentiation, which belongs to usage guidelines. Combined with schema and output schema, an agent has everything needed to invoke it correctly.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema has 0% description coverage, and the description fills in all of it: repo_path is the path to the repository and must contain pyproject.toml for dependency checks, and min_confidence maps to vulture's 0-100 threshold with a false-positive trade-off. This goes well beyond the parameter names and defaults.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description opens with a specific verb-resource pair: scan a Python repository, and names the two tools (vulture, deptry) and the two classes of findings (dead code, dependency issues). This clearly distinguishes it from the sibling propose_removal_pr, whose action is proposing a PR, not scanning.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
No explicit guidance on when to use this tool versus propose_removal_pr, nor any when-not conditions. However, the read-only declaration and human-readable output imply this is an analysis step, and the requirement about pyproject.toml provides some context. This makes usage predictable but not explicitly ruled for alternatives.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
v0.1.2propose_removal_prscan_dead_codeScored across 2 tools
scan_dead_code is purely read-only analysis, while propose_removal_pr actually creates a branch and PR. Their purposes are complementary and unlikely to be confused.
Both tool names use lower_snake verb-first conventions: scan_dead_code and propose_removal_pr. The pattern is predictable, though the second name is slightly less clean because 'removal_pr' bundles an action and an object into one noun phrase.
Two tools is on the thin side and sits at the borderline of feeling complete. However, each tool has a distinct role in a focused analyze-then-propose workflow, so the small count is plausible for a narrow server.
The scan surfaces both dead code and dependency issues, but propose_removal_pr only handles unused imports. Dependency fixes and non-import dead-code removals are therefore dead ends, creating a notable gap in the advertised safe-fix domain.
Unclaimed servers have limited discoverability.
If you are the server author, to access and configure the admin panel.
Codebase intelligence for AI agents — dead code, blast radius, ownership.
Screens public GitHub repos and PRs to generate risk maps, findings, and merge-readiness signals.
Security + bug + perf + refactor audit for Python. Returns 0-10 score + MD report.
Audit and repair any public GitHub repo. Free survey; paid source, unified diffs and a pull request.
