# callgraph (mcpb · callgraph.mcpb)

AST call graph and code map for AI coding agents: PHP, JS, TS, Vue, Rust, Swift, Python, Go.

- Trust score: 43/100 (low)
- Change this week: +3
- Registry status: active
- Liveness: live
- Owner verified: no
- Last scored: 2026-11-22

## Components

- mcpb · `callgraph.mcpb`: 43/100 (this document), [markdown](https://verifymcp.io/servers/geriy-callgraph/callgraph.md), [page](https://verifymcp.io/servers/geriy-callgraph/callgraph)

## Channel facts

- Registry: `mcpb`
- Package: `https://github.com/GeRiY/callgraph/releases/download/v1.1.1/callgraph.mcpb`
- Transport: `stdio`
- SHA-256: `711a9e9aead163edfc6baf91ab72930c365809ab293e115dd36f0f031fd25179`

## Trust breakdown

How this component scores in each security and reliability category. Every signal is checked automatically from public evidence about the published package, including repeated runs of it in an isolated sandbox, and we only credit what we can confirm. Scores are 0–100 per category. Scoring method: https://verifymcp.io/docs/scoring (what has changed: https://verifymcp.io/docs/scoring/changelog)

Scored 2026-11-22.

- **Supply Chain Security**: 0/100
  - No malware scan is available for this kind of package: the supply-chain vendors we use do not cover it. This is a permanent gap in our coverage, not a finding about the package.
  - Known CVEs could not be checked: we found no dependency list to check in this bundle: either it runs a compiled binary, whose bundled manifests we do not read, or none of the package.json, requirements.txt, pyproject.toml or uv.lock files it ships for its runtime declares a dependency we can read (a requirements.txt that only points at a setup.py project, a local path, or a nested requirements file we cannot follow counts as none).
  - Install-script risk not yet assessed.
  - Dependency health could not be checked: we found no dependency list to check in this bundle: either it runs a compiled binary, whose bundled manifests we do not read, or none of the package.json, requirements.txt, pyproject.toml or uv.lock files it ships for its runtime declares a dependency we can read (a requirements.txt that only points at a setup.py project, a local path, or a nested requirements file we cannot follow counts as none).
- **Provenance & Transparency**: 48/100
  - Source repository is publicly reachable at the declared URL.
  - Provenance check failed: no build-provenance attestation is published.
  - Clear OSI-approved license (MIT).
  - Actively maintained (last published 3 days ago).
  - Publishes a security disclosure policy (SECURITY.md).
- **Schema Quality & AI Usability**: 58/100
  - AI-judged instruction clarity (excellent).
  - Context-footprint check failed: tool/resource definitions use about 4248 tokens (~354/item across 12 items; 12 tools + 0 resources), over budget; trim descriptions and params.
  - Usage-examples check failed: none of the tools include examples.
- **Stability & Change Management**: 33/100
  - Stability observed for 10 of 30 days with no destabilising changes; credit accrues until the full window elapses.
- **Tool Coverage**: 100/100
  - 100% of tools have a non-trivial description (not blank, and not just the tool's name).
  - 100% of tool parameters carry a description.
- **Tool Safety**: 100/100
  - No prompt-injection markers were found in the server instructions, tool names or descriptions we captured.
  - We read all 12 captured tool definition(s), and no name or description among them implies an irreversible operation.
  - An AI judge read all 13 captured unit(s) of tool text and found none that tries to manipulate the model reading it.
- **Capabilities**: 100/100
  - Implements a current MCP spec version (2026-07-28).

**Unverified: 1 category.** A category scored 0 because we could not verify it: a data source with nothing on this package, evidence we could not reach, or a check we could not run. We only credit what we can confirm.

## Install

### How do I install the callgraph MCP server?

callgraph ships as an MCPB bundle, a single file you download and open in an MCP host that supports MCPB bundles, such as Claude Desktop. The host reads the launch command from the bundle's own manifest, so there is no command to copy. Before opening it, check the file's SHA-256 matches the one the MCP registry declares for it, 711a9e9aead163edfc6baf91ab72930c365809ab293e115dd36f0f031fd25179; the commands to do that are on this page. If it does not match, do not open it.

- Download bundle: `https://github.com/GeRiY/callgraph/releases/download/v1.1.1/callgraph.mcpb`
- SHA-256: `711a9e9aead163edfc6baf91ab72930c365809ab293e115dd36f0f031fd25179`

### Verify on macOS / Linux

```bash
echo "711a9e9aead163edfc6baf91ab72930c365809ab293e115dd36f0f031fd25179  callgraph.mcpb" | shasum -a 256 -c
```

### Verify on Windows (PowerShell)

```powershell
(Get-FileHash -Algorithm SHA256 .\callgraph.mcpb).Hash -eq "711a9e9aead163edfc6baf91ab72930c365809ab293e115dd36f0f031fd25179"
```

## Changelog

Every change recorded for this component, newest first. Days that predate change tracking, or that we cannot explain, say so: "we were watching and nothing happened" and "we were not watching" are different claims.

### 2026-11-22 (score 43, +1)

No change was recorded against any check on this day. Stability & Change Management went from 30 to 33. That category is still filling its 30-day observation window: 9 days of observed history at the previous scan, 10 at this one. The score rises as the window fills, whether or not the server changes.

### 2026-11-20 (score 42, +1)

No change was recorded against any check on this day. Stability & Change Management went from 23 to 27. That category is still filling its 30-day observation window: 7 days of observed history at the previous scan, 8 at this one. The score rises as the window fills, whether or not the server changes.

### 2026-11-18 (score 41, +1)

No change was recorded against any check on this day. Stability & Change Management went from 17 to 20. That category is still filling its 30-day observation window: 5 days of observed history at the previous scan, 6 at this one. The score rises as the window fills, whether or not the server changes.

### 2026-11-15 (score 40, +1)

No change was recorded against any check on this day. Stability & Change Management went from 7 to 10. That category is still filling its 30-day observation window: 2 days of observed history at the previous scan, 3 at this one. The score rises as the window fills, whether or not the server changes.

### 2026-11-13 (score 39, +1)

- [functional improvement] Stability: unverified → 0.03

### 2026-11-12 (score 38)

First indexed and scored.

## MCP tools (12)

### `find_callers` (~468 tokens)

Who calls it: each caller's file, method, line range and call line (`L25 │ code`); paths relative to the `root:` line. A plain word (`cancel`) matches the whole name (`x.cancel`, `T::cancel`), not `cancelAll`; regex chars, `.` or `::` keep regex substring matching (`.*cancel.*` = substring). Name regex `(^|::|\.|->)cancel$`: every call, also with an unknown/inherited receiver. `Order::cancel` = only calls resolved to that type (`via <Type>` = through interface/parent/trait), so it can miss callers; `?::name` = name-only match. `depth` 2–5 = impact analysis (callers of callers), `compact: true` = file-grouped names + lines, no code; `guess` adds name-only matches (?). Example: {"pattern": "(^|::|\\.|->)cancel$", "path": "app"}.

Input parameters:

- `by_layer` (boolean): Group callers by layer (closest to the callee first).
- `compact` (boolean): With `depth` 2–5: file-grouped `Type::name()  L<start>` lines without code (far fewer tokens).
- `depth` (integer|null): 1 (default) = direct callers; 2–5 = callers of callers (impact analysis).
- `full` (boolean): Disable the 400-line output cap.
- `guess` (boolean): Also follow name-only matches (marked `?`).
- `layer` (string|null): Search only files of this layer (level 1), e.g. `controller`.
- `path` (string|null): Search scope (file or directory) relative to the project root; default: the whole root.
- `pattern` (string, required): Callee: a plain word `cancel` = the whole name; regex `(^|::|\.|->)cancel$`, `.*cancel.*` (substring); `Order::cancel` (resolved to that type); JS keys: `useForm.js::submit`, `route::orders.show`.
- `project` (string|null): Registered project alias; default: the server's working directory.

### `framework_view` (~252 tokens)

Framework wiring of the directory (or file) holding the definitions: which route → which controller, which event → which listener; also middleware, validation, jobs, providers, magic, di, views; inertia: page ← render site, props; ziggy: JS route('name') calls checked against the Laravel route table; forms: Inertia useForm fields checked against the controller validation rules; models: Eloquent models with their table columns (from migrations), fillable/casts and relations; i18n: translation keys missing per locale, unused keys, and $t()/__() calls whose key does not exist (a lang file with ::key shows where the key is used); classes: stylesheet classes and which templates use them, plus classes nothing uses. Example: {"view": "routes", "path": "routes"}. Structure of a file or folder: `outline`.

Input parameters:

- `full` (boolean): Disable the 400-line output cap.
- `path` (string, required): Directory (or file) relative to the project root, e.g. `routes`, `app`.
- `project` (string|null): Registered project alias; default: the server's working directory.
- `view` (required): Which view.

### `guide` (~130 tokens)

The core callgraph manual (SKILL.md): queries with CLI examples, formats, output limit. Call it once when the server instructions are not enough. `topic` returns one reference file instead: php, blade, js, jsx, vue, css, rust, swift, kotlin, python, go, gen, mcp, reference. Example: {"topic": "php"}.

Input parameters:

- `topic` (string|null): One reference file instead of the core manual: php, blade, js, jsx, vue, css, rust, swift, kotlin, python, go, gen, mcp or reference.

### `locate` (~290 tokens)

Where to start with a task: the task's names and domain terms in ONE call → a ranked file list from the project map (no file reading; top 5, `all: true` = all), each with reasons (exact definition, file name, name part, string literal, call, `callers: N`). Example: {"terms": ["RefundController", "refund", "payment_status"]}. Prefer code names; a plain sentence also works (filler words are ignored); translate non-English domain words (visszatérítés → refund); one call with all terms. Use `locate` to find WHERE to start; once you know the file use `outline`; string literal / constant: `search`; other exact text (log message, config value): Grep.

Input parameters:

- `all` (boolean): Every matching file instead of the top 5.
- `project` (string|null): Registered project alias; default: the server's working directory.
- `terms` (required): The task's names and terms in code form (function / class / file names, columns, routes, English code names of non-English words); a string works too. An element with spaces = one phrase matched agai…

### `method_body` (~219 tokens)

Line-numbered source of methods/functions, compact: docblock and blank lines dropped, comments in the body kept (line numbers stay exact; the header notes an omitted docblock; `doc: true` adds it, `raw: true` = exact text, e.g. before editing). Use after `outline` instead of reading the whole file. Example: {"path": "app/Order.php", "method": ["cancel", "refund"]}.

Input parameters:

- `doc` (boolean): Also include the docblock / doc comments above each method.
- `full` (boolean): Disable the 400-line output cap.
- `method` (required): Method/function name (`"cancel"`), or an array of names (`["cancel", "refund"]`).
- `path` (string, required): File path, relative to the project root.
- `project` (string|null): Registered project alias; default: the server's working directory.
- `raw` (boolean): Exact, unfiltered source (docblock and blank lines kept), e.g. before an exact edit.

### `next` (~179 tokens)

Run a suggested next step by number instead of composing the call: some results end with `↪ r7 possible next: [1] … [2] …` (server run with `--next`). Use it only if a suggestion fits your task; else ignore the line. Example: {"result_id": "r7", "choice": [1, 3]}. Ids are per project and expire after 24 hours.

Input parameters:

- `choice` (required): Which suggestion(s): `1`, `"1"`, `"1,3"` or `[1, 3]`.
- `project` (string|null): Registered project alias (the one the result came from); default: the server's working directory.
- `result_id` (string, required): The result id from the `↪ rN …` line, e.g. `"r7"`.

### `notes` (~155 tokens)

Notes (pilot) on files or methods by people/agents: `path[::method]  time  📝 text`, [orphaned] if the method is gone. Read-only (written with the CLI `callgraph note add`). Example: {"pattern": "refund", "path": "app"}; no `pattern`: all notes.

Input parameters:

- `path` (string|null): Only notes under this file or directory (relative to the project root); default: the whole root.
- `pattern` (string|null): Regex on the note text or its `path::method` anchor (case-insensitive); default: all notes.
- `project` (string|null): Registered project alias; default: the server's working directory.

### `outline` (~306 tokens)

Structure of a code file or directory without reading it: classes/components, methods with line ranges, calls by category. With `method`: that method's callees, each classified (internal_helper, model_use, orm_relation, framework_use, external_call…) — no separate callees tool; who calls it: `find_callers`. Start here once you know the file (else `locate`); routes / events / validation: `framework_view`. Example: {"path": "app/Order.php", "method": "cancel"}. Pitfall: the method goes in `method`, not `path::method`; on a large directory use `format: summary` or a subdirectory.

Input parameters:

- `format`: outline (default): classes, methods with line ranges, calls by category; api: public surface; flow: control branches; summary: one line per file; hotspots: most-called; stats: counts.
- `full` (boolean): Disable the 400-line output cap (use only when needed).
- `layer` (string|null): Only files of this layer (see `project_profile`), e.g. `service`, `service,repository`, or one ecosystem: `controller:js`.
- `method` (string|null): Only this method/function (file paths only).
- `path` (string, required): File or directory, relative to the project root.
- `project` (string|null): Registered project alias (see `projects`); default: the server's working directory.

### `package_usage` (~127 tokens)

Where and how the code uses a composer / npm dependency: per API symbol the count, first file and +N; `calls: true` lists every call per file. Use before upgrading or replacing a package. Example: {"package": "guzzle"}.

Input parameters:

- `calls` (boolean): Every call line per file instead of the per-symbol summary.
- `package` (string, required): Package name or short name: `guzzlehttp/guzzle`, `guzzle`, `@vueuse/core`.
- `project` (string|null): Registered project alias; default: the server's working directory.

### `project_profile` (~241 tokens)

Project profile from folder and file names, checked against the code graph (~15 lines): architecture (Laravel/Yii MVC, DDD/hexagonal, Vue/React, SwiftUI MVVM, Rust web), layers with folders and file counts, patterns, call direction between layers, unknown folders, notable packages with locked versions (`stack:`); `packages: true` lists every direct composer / npm dependency. Call it first in an unfamiliar project, then pass `layer` (e.g. "service") to `outline` / `search` / `find_callers`. Example: {} or {"full": true}. Pitfall: a layer name comes from this list (controller, service, repository, model, component, store…), not a folder path.

Input parameters:

- `deps` (boolean): Instead: probably unused, missing and dev-only composer / npm dependencies.
- `full` (boolean): Per-folder details, confidence counts and the call pairs between layers.
- `packages` (boolean): Also list every direct composer / npm dependency: installed and declared version, dev flag.
- `project` (string|null): Registered project alias; default: the server's working directory.

### `projects` (~49 tokens)

Registered project aliases and their roots (for the `project` parameter). Call it when the server's working directory is not the project (e.g. Claude Desktop), then pass `project` to every tool.

### `search` (~176 tokens)

Where a string literal (error message, key) or a constant is used/defined (substring match). Example: {"kind": "strings", "pattern": "Order not found"}. Calls of a function: `find_callers`; which files to start a task in: `locate`, not this.

Input parameters:

- `full` (boolean): Disable the 400-line output cap.
- `kind` (required): strings: string literals (messages, keys); constants: constant references and definitions.
- `layer` (string|null): Only files of this layer, e.g. `service`.
- `path` (string|null): Search scope relative to the project root; default: the whole root.
- `pattern` (string, required): Text to look for (substring).
- `project` (string|null): Registered project alias; default: the server's working directory.

## Diagnostics

Captured diagnostic sections: Provenance. The full working is on the page: https://verifymcp.io/servers/geriy-callgraph/callgraph#diagnostics

## Score history

- 2026-11-22: 43
- 2026-11-21: 42
- 2026-11-20: 42
- 2026-11-19: 41
- 2026-11-18: 41
- 2026-11-17: 40
- 2026-11-16: 40
- 2026-11-15: 40
- 2026-11-14: 39
- 2026-11-13: 39
- 2026-11-12: 38

## Common questions

### What is the callgraph MCP server?

callgraph is an MCP server listed in the public MCP registry as io.github.GeRiY/callgraph. AST call graph and code map for AI coding agents: PHP, JS, TS, Vue, Rust, Swift, Python, Go. This page covers its MCPB bundle (https://github.com/GeRiY/callgraph/releases/download/v1.1.1/callgraph.mcpb).

### Is the callgraph MCP server safe to use?

callgraph scores 43 out of 100 on VerifyMCP. That is a record of what we were able to check automatically, not an endorsement. The category breakdown on this page shows every signal behind the number, including the ones we could not confirm.

### What tools does the callgraph MCP server expose?

callgraph exposes 12 tools: find_callers, framework_view, guide, locate, method_body, and 7 more. Their descriptions and schemas cost roughly 2,592 tokens of context every time the server is loaded.

### What licence is the callgraph MCP server under?

callgraph declares the MIT licence, which is OSI-approved. That covers the source only, and says nothing about the cost of any service it calls.

### What did VerifyMCP test on the callgraph MCP server?

VerifyMCP ran 12 checks on callgraph, covering Source repository, Provenance, License, Maintenance and 8 more. 4 items could not be completed, a gap on our side rather than a finding about callgraph: Malware scan, Known CVEs, Install scripts and Dependency health. One category scores 0 for that reason, because we only credit what we can confirm. Endpoint Security and Transport & Reachability do not apply to an MCPB bundle, so they were not tested. The score does not measure everything, and the scoring documentation lists what it leaves out.

## Links

- Repository: https://github.com/GeRiY/callgraph
- Website: https://callgraph.eu/
- Changelog RSS feed: https://verifymcp.io/servers/geriy-callgraph/callgraph.xml
- Changelog JSON feed: https://verifymcp.io/servers/geriy-callgraph/callgraph/changelog.json
- Trust report (JSON): https://verifymcp.io/servers/geriy-callgraph/callgraph.json
- HTML version of this page: https://verifymcp.io/servers/geriy-callgraph/callgraph
