# chunk-invariance (pypi · chunk-invariance)

Checks whether your streaming redactor leaks a secret split across chunks; writes a pytest for it.

- Trust score: 76/100 (medium)
- Change this week: +3
- Registry status: active
- Liveness: live
- Owner verified: no
- Last scored: 2026-11-05

## Components

- pypi · `chunk-invariance`: 76/100 (this document), [markdown](https://verifymcp.io/servers/ninadphalak-chunk-invariance/chunk-invariance.md), [page](https://verifymcp.io/servers/ninadphalak-chunk-invariance/chunk-invariance)

## Channel facts

- Registry: `pypi`
- Package: `chunk-invariance`
- Version: `0.2.0`
- Transport: `stdio`

## 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 2026-11-05.

- **Supply Chain Security**: 100/100
  - No malware found by supply-chain analysis.
  - No known CVEs affecting this package version or its production dependencies.
  - Runs setuptools.build_meta at install time, a recognised build step with no custom scripting around it.
  - No production dependencies, so there is no dependency health to assess.
- **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 (Apache-2.0).
  - Actively maintained (last published 17 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 1565 tokens (~521/item across 3 items; 3 tools + 0 resources), over budget; trim descriptions and params.
  - Usage-examples check failed: none of the tools include examples.
- **Stability & Change Management**: 57/100
  - Stability observed for 17 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 3 captured tool definition(s), and no name or description among them implies an irreversible operation.
  - An AI judge read all 4 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).

## Install

### How do I install the chunk-invariance MCP server?

chunk-invariance runs locally as a PyPI package, launched with uvx chunk-invariance. Ready-made configuration for Claude, Cursor, VS Code, Codex and 5 more is on this page, copied from each client's own documentation.

### Claude

```bash
claude mcp add ninadphalak-chunk-invariance -- uvx chunk-invariance
```

### Cursor

```json
{
  "mcpServers": {
    "ninadphalak-chunk-invariance": {
      "command": "uvx",
      "args": [
        "chunk-invariance"
      ]
    }
  }
}
```

### VS Code

```json
{
  "servers": {
    "ninadphalak-chunk-invariance": {
      "command": "uvx",
      "args": [
        "chunk-invariance"
      ]
    }
  }
}
```

### Codex

```bash
codex mcp add ninadphalak-chunk-invariance -- uvx chunk-invariance
```

### opencode

```json
{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "ninadphalak-chunk-invariance": {
      "type": "local",
      "command": [
        "uvx",
        "chunk-invariance"
      ],
      "enabled": true
    }
  }
}
```

### OpenClaw

```bash
openclaw mcp add ninadphalak-chunk-invariance --command uvx --arg chunk-invariance
```

### Hermes

```yaml
mcp_servers:
  ninadphalak-chunk-invariance:
    command: "uvx"
    args: ["chunk-invariance"]
```

### Netclaw

```json
{
  "McpServers": {
    "ninadphalak-chunk-invariance": {
      "Transport": "stdio",
      "Command": "uvx",
      "Arguments": [
        "chunk-invariance"
      ]
    }
  }
}
```

### Vellum

```bash
assistant mcp add ninadphalak-chunk-invariance -t stdio -c uvx -a chunk-invariance
```

### Other

```json
{
  "mcpServers": {
    "ninadphalak-chunk-invariance": {
      "command": "uvx",
      "args": [
        "chunk-invariance"
      ]
    }
  }
}
```

## 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.

### 2026-11-05 (score 76, +16)

- [security improvement] Malware scan: unverified → pass

### 2026-11-04 (score 60, −15)

- [security regression] Malware scan: pass → unverified

### 2026-11-02 (score 75, +16)

- [security improvement] Malware scan: unverified → pass

### 2026-11-01 (score 59, −15)

- [security regression] Malware scan: pass → unverified

### 2026-10-31 (score 74, +1)

No change was recorded against any check on this day. Stability & Change Management went from 37 to 40. That category is still filling its 30-day observation window: 11 days of observed history at the previous scan, 12 at this one. The score rises as the window fills, whether or not the server changes.

### 2026-10-29 (score 73, +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.

### 2026-10-27 (score 72, +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.

### 2026-10-25 (score 71, +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.

## MCP tools (3)

### `check_stream_filter` (~673 tokens)

Check whether the user's streaming filter leaks or changes values that are split across chunks. Streams each sample through the filter in every two-part split (or more parts) and compares the joined output with filtering the whole text at once. Returns PASS or FAIL, the first failing split in plain words (which value came out unredacted), the number of splits checked, and how to fix it. It imports and runs the user's own code from project_dir on this machine, in a separate process that is stopped after timeout_seconds. The server makes no network calls and writes no files. Its result repeats the samples, any value the filter leaked and the message of any exception the user's code raised, and that result goes into this conversation, so use synthetic samples only, never a real secret or real personal data. What the user's code prints is never returned.

Input parameters:

- `flush_target`: per_chunk only: package.module:function for flush(state), called once after the last chunk.
- `project_dir` (string, required): Absolute path to the user's project root. The target is imported from here (and from its src/ folder).
- `python`: Interpreter that imports the user's code. Default: project_dir/.venv or project_dir/venv if present, else the server's own Python.
- `samples`: Input texts that contain values the filter should redact or rewrite. Default: synthetic samples (an email, an SSN, a card number, API-key shapes; all fake): ['Contact jane.doe@example.com about the i…
- `shape` (string): generator: target(chunks) takes an iterable of text chunks and yields or returns output chunks. per_chunk: target(chunk, state) returns the output for one chunk; state is a dict kept across the strea…
- `splits`: 'all-two-part' (default): the whole text plus every single cut. An int up to 3: every split into at most that many parts (slower).
- `target` (string, required): The streaming filter as package.module:function, importable from project_dir the way the user's tests import it.
- `timeout_seconds` (number): Seconds before the check process is stopped.
- `whole_target`: The non-streaming function, package.module:function, taking the whole text and returning the filtered text. Default: the stream filter run on the whole text as one chunk, which cannot catch a missing…

### `write_regression_test` (~682 tokens)

Write a pytest file that runs chunk_invariance.assert_chunk_invariant on the user's streaming filter for each sample, so their CI keeps checking that values split across chunks cannot leak. Takes the same inputs as check_stream_filter, first checks that the target imports, and returns the file text with a suggested file name and whether it passes today. Save it in the user's tests folder; the test environment needs `pip install chunk-invariance`. It imports and runs the user's own code from project_dir on this machine, in a separate process that is stopped after timeout_seconds. The server makes no network calls and writes no files. Its result repeats the samples, any value the filter leaked and the message of any exception the user's code raised, and that result goes into this conversation, so use synthetic samples only, never a real secret or real personal data. What the user's code prints is never returned.

Input parameters:

- `flush_target`: per_chunk only: package.module:function for flush(state), called once after the last chunk.
- `project_dir` (string, required): Absolute path to the user's project root. The target is imported from here (and from its src/ folder).
- `python`: Interpreter that imports the user's code. Default: project_dir/.venv or project_dir/venv if present, else the server's own Python.
- `samples`: Input texts that contain values the filter should redact or rewrite. Default: synthetic samples (an email, an SSN, a card number, API-key shapes; all fake): ['Contact jane.doe@example.com about the i…
- `shape` (string): generator: target(chunks) takes an iterable of text chunks and yields or returns output chunks. per_chunk: target(chunk, state) returns the output for one chunk; state is a dict kept across the strea…
- `splits`: 'all-two-part' (default): the whole text plus every single cut. An int up to 3: every split into at most that many parts (slower).
- `target` (string, required): The streaming filter as package.module:function, importable from project_dir the way the user's tests import it.
- `timeout_seconds` (number): Seconds before the check process is stopped.
- `whole_target`: The non-streaming function, package.module:function, taking the whole text and returning the filtered text. Default: the stream filter run on the whole text as one chunk, which cannot catch a missing…

### `explain` (~34 tokens)

Explain in one paragraph why filtering a stream chunk by chunk leaks values cut across chunks, with an example and the fix. No inputs.

## Diagnostics

Captured diagnostic sections: Provenance, Install scripts, Dependencies. The full working is on the page: https://verifymcp.io/servers/ninadphalak-chunk-invariance/chunk-invariance#diagnostics

## Score history

- 2026-11-05: 76
- 2026-11-04: 60
- 2026-11-03: 75
- 2026-11-02: 75
- 2026-11-01: 59
- 2026-10-31: 74
- 2026-10-30: 73
- 2026-10-29: 73
- 2026-10-28: 72
- 2026-10-27: 72
- 2026-10-26: 71
- 2026-10-25: 71
- 2026-10-24: 70
- 2026-10-23: 70
- 2026-10-22: 69
- 2026-10-21: 69
- 2026-10-20: 68
- 2026-10-19: 68

## Common questions

### What is the chunk-invariance MCP server?

chunk-invariance is an MCP server listed in the public MCP registry as io.github.ninadphalak/chunk-invariance. Checks whether your streaming redactor leaks a secret split across chunks; writes a pytest for it. This page covers its PyPI package (chunk-invariance).

### Is the chunk-invariance MCP server safe to use?

chunk-invariance scores 76 out of 100 on VerifyMCP. We found no known CVEs affecting it as of 5 November 2026. 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 chunk-invariance MCP server expose?

chunk-invariance exposes 3 tools: check_stream_filter, write_regression_test, explain. Their descriptions and schemas cost roughly 1,389 tokens of context every time the server is loaded.

### Is the chunk-invariance MCP server still maintained?

chunk-invariance is still listed as active in the MCP registry. We last reached this channel on 5 November 2026. Those dates come from our own scans of the registry and the channel itself, not from anything the publisher announced.

### What licence is the chunk-invariance MCP server under?

chunk-invariance declares the Apache-2.0 licence, which is OSI-approved. That covers the source only, and says nothing about the cost of any service it calls.

## Links

- PyPI project: https://pypi.org/project/chunk-invariance/
- Socket report: https://socket.dev/pypi/package/chunk-invariance
- Repository: https://github.com/ninadphalak/LLM-Shield-Proxy
- Website: https://github.com/ninadphalak/LLM-Shield-Proxy/tree/main/chunk-invariance
- Changelog RSS feed: https://verifymcp.io/servers/ninadphalak-chunk-invariance/chunk-invariance.xml
- Changelog JSON feed: https://verifymcp.io/servers/ninadphalak-chunk-invariance/chunk-invariance.json
- HTML version of this page: https://verifymcp.io/servers/ninadphalak-chunk-invariance/chunk-invariance