# TokenMark (npm · @altronis/tokenmark-mcp)

Measured local-LLM speeds and model picks for Strix Halo, DGX Spark, Mac and GPU cards.

- Trust score: 52/100 (low)
- Change this week: −11
- Registry status: active
- Liveness: live
- Owner verified: no
- Last scored: 2026-10-23

## Components

- npm · `@altronis/tokenmark-mcp`: 52/100 (this document), [markdown](https://verifymcp.io/servers/sypherin-tokenmark-mcp/altronis-tokenmark-mcp.md), [page](https://verifymcp.io/servers/sypherin-tokenmark-mcp/altronis-tokenmark-mcp)

## Channel facts

- Registry: `npm`
- Package: `@altronis/tokenmark-mcp`
- Version: `0.4.1`
- 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 2026-10-23.

- **Supply Chain Security**: 50/100
  - Malware scan not yet available for this package.
  - No known CVEs affecting this package version or its production dependencies.
  - No install/post-install scripts declared.
  - 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 (MIT).
  - Actively maintained (last published 4 days ago).
  - Publishes a security disclosure policy (SECURITY.md).
- **Schema Quality & AI Usability**: 60/100
  - AI-judged instruction clarity (good).
  - Context-footprint check failed: tool/resource definitions use about 1477 tokens (~211/item across 7 items; 7 tools + 0 resources), over budget; trim descriptions and params.
  - Usage-examples check failed: none of the tools include examples.
- **Stability & Change Management**: 30/100
  - Stability observed for 9 of 30 days with no destabilising changes; credit accrues until the full window elapses.
- **Tool Coverage**: 91/100
  - 100% of tools have a non-trivial description (not blank, and not just the tool's name).
  - 72% of tool parameters carry a description.
- **Tool Safety**: 75/100
  - No prompt-injection markers were found in the server instructions, tool names or descriptions we captured.
  - 0 of 1 tool(s) whose name or description implies an irreversible operation declare an MCP destructiveHint annotation; "tokenmark_submit_result" implies "send" and declares no destructiveHint at all, which the MCP spec reads as destructive by default.
  - An AI judge read all 7 captured unit(s) of tool text and found none that tries to manipulate the model reading it.
- **Capabilities**: 20/100
  - Spec-recency check failed: implements MCP spec 2024-11-05; the latest is 2026-07-28.

## Install

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

TokenMark runs locally as an npm package, launched with npx -y @altronis/tokenmark-mcp. 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 sypherin-tokenmark-mcp -- npx -y @altronis/tokenmark-mcp
```

### Cursor

```json
{
  "mcpServers": {
    "sypherin-tokenmark-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "@altronis/tokenmark-mcp"
      ]
    }
  }
}
```

### VS Code

```json
{
  "servers": {
    "sypherin-tokenmark-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "@altronis/tokenmark-mcp"
      ]
    }
  }
}
```

### Codex

```bash
codex mcp add sypherin-tokenmark-mcp -- npx -y @altronis/tokenmark-mcp
```

### opencode

```json
{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "sypherin-tokenmark-mcp": {
      "type": "local",
      "command": [
        "npx",
        "-y",
        "@altronis/tokenmark-mcp"
      ],
      "enabled": true
    }
  }
}
```

### OpenClaw

```bash
openclaw mcp add sypherin-tokenmark-mcp --command npx --arg -y --arg @altronis/tokenmark-mcp
```

### Hermes

```yaml
mcp_servers:
  sypherin-tokenmark-mcp:
    command: "npx"
    args: ["-y", "@altronis/tokenmark-mcp"]
```

### Netclaw

```json
{
  "McpServers": {
    "sypherin-tokenmark-mcp": {
      "Transport": "stdio",
      "Command": "npx",
      "Arguments": [
        "-y",
        "@altronis/tokenmark-mcp"
      ]
    }
  }
}
```

### Vellum

```bash
assistant mcp add sypherin-tokenmark-mcp -t stdio -c npx -a -y @altronis/tokenmark-mcp
```

### Other

```json
{
  "mcpServers": {
    "sypherin-tokenmark-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "@altronis/tokenmark-mcp"
      ]
    }
  }
}
```

## 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.

### 2026-10-23 (score 52, −15)

- [security regression] Malware scan: pass → unverified
- [functional regression] Schema quality: 173 → 211
- [functional] Package version: 0.3.0 → 0.4.1

### 2026-10-22 (score 67, +4)

- [functional improvement] Stability: unverified → 0.27

### 2026-10-14 (score 63, +23)

- [security improvement] Injection markers: unverified → pass
- [security] First check of Judged manipulation: pass
- [security] Stability: Stability not yet verified: not enough scan history yet (needs a 30-day window).
- [functional regression] MCP protocol: unverified → fail
- [functional improvement] Tool coverage: unverified → 100
- [functional] First check of Schema quality: fail
- [functional] First check of Schema quality: fail
- [functional] First check of Tool coverage: 71
- [functional] First check of Destructive annotations: 0
- [functional] First check of Schema quality: good

### 2026-10-13 (score 40, +1)

- [functional improvement] Security disclosure: unverified → pass

### 2026-10-12 (score 39, −1)

- [functional regression] Security disclosure: pass → unverified
- [functional] Package version: 0.2.1 → 0.3.0

### 2026-10-06 (score 40)

First indexed and scored.

## MCP tools (7)

### `tokenmark_recommend` (~257 tokens)

Recommend the best local-LLM model + config for given hardware and tasks, from real tracked benchmarks. Returns ranked picks with measured tok/s and a source link. Never invents numbers. config.speculative=true means the tok/s came from speculative decoding (MTP, DFlash, a draft model), not a plain run of the same quant; say so when quoting it. AI boxes (unified memory, up to 128 GB+) and GPU cards (own VRAM) are separate lanes: `lane` says which one the picks came from; never present a card pick and a box pick as one ranking.

Input parameters:

- `cards` (boolean): true: read the GPU-card lane (RTX 5090, RTX PRO 6000) instead of the AI boxes. Naming a card as hardware does the same. The two lanes are never ranked together.
- `hardware` (string, required): e.g. "Strix Halo", "Gorgon Halo", "DGX Spark", "Mac Ultra", "RTX 5090"
- `limit` (number): how many picks (default 5, max 10)
- `prefer` (string): default balanced
- `tasks` (array): what the user wants to do

### `tokenmark_configs` (~180 tokens)

List tracked benchmark configs, optionally filtered by model and/or hardware. Returns measured decode tok/s, quant, backend, run mode (speculative=true for MTP/DFlash/draft-model runs), author, and source. Reads one lane: the AI boxes, or the GPU cards when cards=true or hardware names a card; `other_lane` counts what the same filters match in the other lane.

Input parameters:

- `cards` (boolean): true: read the GPU-card lane (RTX 5090, RTX PRO 6000) instead of the AI boxes. Naming a card as hardware does the same. The two lanes are never ranked together.
- `hardware` (string)
- `limit` (number): rows to return, fastest first (default 25, max 50); `matched` gives the full count
- `model` (string)

### `tokenmark_submit` (~185 tokens)

Submit a GitHub or Hugging Face repo whose README/model card contains local-LLM benchmark numbers (tok/s on Strix Halo, DGX Spark, Mac, etc.) to TokenMark. Configs are extracted on the tracker, then a human reviews before anything goes live. Returns a submission id to poll with tokenmark_submission_status. A speed the user measured themselves (no repo) goes through tokenmark_submit_result; hardware the catalogue is missing goes through the signed-in form at https://tokenmark.app/submit.

Input parameters:

- `note` (string): optional context for the reviewer (hardware, backend build, how it was measured)
- `repo` (string, required): https://github.com/owner/name, https://huggingface.co/owner/name, or owner/name
- `source` (string): only needed for a bare owner/name that is on Hugging Face

### `tokenmark_submit_result` (~531 tokens)

Send a local-LLM speed the user measured themselves to TokenMark's review queue, under their account. Pass raw llama-bench output (-o json, jsonl, csv or the markdown table) or vllm bench serve output as bench_output and it is read directly; typed fields override what it says. Needs a one-time sign-in on this machine: the user runs `npx @altronis/tokenmark-cli login` in a terminal and approves the code in their browser (or TOKENMARK_TOKEN is set). Call it with dry_run=true first: that returns a one-line summary and a confirm code. Show the user the summary and get their ok, then call again with the same fields and confirm set to that code; sending without it is refused. Never send a number the user did not measure. A person reviews every result before it goes live. Returns a submission id for tokenmark_submission_status.

Input parameters:

- `backend` (string): e.g. llama.cpp, vllm, mlx, ollama
- `backend_version` (string)
- `bench_output` (string): raw stdout of llama-bench or vllm bench serve
- `command` (string): the exact command that was run
- `concurrency` (number)
- `confirm` (string): the confirm code from the dry run, once the user has said the summary is right
- `credit` (string): handle to credit
- `ctx` (number): context length in tokens
- `decode_tps` (number): decode tokens/s (0.1-5000)
- `device` (string): the exact box, e.g. an OEM model
- `dry_run` (boolean): true: check and summarise without sending; returns the confirm code
- `hardware` (string): e.g. "Strix Halo", "DGX Spark", "strix-halo", or any device name (kept as other)
- `mode` (string): speculative-decoding mode, if used
- `model` (string): e.g. Qwen3.8-27B
- `note` (string)
- `pick` (number): which setup to send when bench_output has more than one (1-based; the error lists them)
- `prefill_tps` (number)
- `proof_url` (string)
- `quant` (string): e.g. Q4_K_M, UD-Q4_K_XL
- `ram_gb` (number)
- `variant` (string): e.g. vulkan, rocm, cuda

### `tokenmark_submission_status` (~50 tokens)

Check a TokenMark submission: queued, extracting, extracted (awaiting review), no_configs, failed, approved or rejected — with the extracted configs and reviewer flags.

Input parameters:

- `id` (string, required)

### `tokenmark_hardware` (~215 tokens)

The TokenMark hardware catalogue: unified-memory platforms for local LLMs (AMD Strix Halo, AMD Gorgon Halo, NVIDIA DGX Spark, Apple M Max / M Ultra), plus the GPU cards with measured runs (gpu_cards: RTX 5090, RTX PRO 6000), which rank in their own lane. Without platform: one row per platform (memory, bandwidth, box count, tracked configs) and one per card. With platform: chip specs with a source link per value, the boxes that ship it, Singapore prices, and what no source states; for a card, its VRAM and bandwidth with sources and its tracked config count. Specs only, no speed numbers; use tokenmark_recommend or tokenmark_configs for tok/s.

Input parameters:

- `platform` (string): id or name, e.g. "gorgon-halo", "Gorgon Halo", "Ryzen AI Max+ 395", "GMKtec EVO-X5 Pro", "Mac Ultra", "RTX 5090"

### `tokenmark_search` (~59 tokens)

Search the tracker for models/configs matching a term (model name, quant, backend, author). AI-box matches are models/matches; GPU-card matches, when there are any, come apart under gpu_cards.

Input parameters:

- `term` (string, required)

## Diagnostics

Captured diagnostic sections: Provenance, Dependencies. The full working is on the page: https://verifymcp.io/servers/sypherin-tokenmark-mcp/altronis-tokenmark-mcp#diagnostics

## Score history

- 2026-10-23: 52
- 2026-10-22: 67
- 2026-10-21: 63
- 2026-10-20: 63
- 2026-10-19: 63
- 2026-10-18: 63
- 2026-10-17: 63
- 2026-10-16: 63
- 2026-10-15: 63
- 2026-10-14: 63
- 2026-10-13: 40
- 2026-10-12: 39
- 2026-10-09: 40
- 2026-10-07: 40
- 2026-10-06: 40

## Common questions

### What is the TokenMark MCP server?

TokenMark is an MCP server listed in the public MCP registry as io.github.sypherin/tokenmark-mcp. Measured local-LLM speeds and model picks for Strix Halo, DGX Spark, Mac and GPU cards. This page covers its npm package (@altronis/tokenmark-mcp).

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

TokenMark scores 52 out of 100 on VerifyMCP. We found no known CVEs affecting it as of 23 October 2026. It declares no install or post-install scripts. 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 TokenMark MCP server expose?

TokenMark exposes 7 tools: tokenmark_recommend, tokenmark_configs, tokenmark_submit, tokenmark_submit_result, tokenmark_submission_status, and 2 more. Their descriptions and schemas cost roughly 1,477 tokens of context every time the server is loaded.

### Is the TokenMark MCP server still maintained?

TokenMark is still listed as active in the MCP registry. We last reached this channel on 23 October 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 TokenMark MCP server under?

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

## Links

- npm package: https://www.npmjs.com/package/@altronis/tokenmark-mcp
- Socket report: https://socket.dev/npm/package/@altronis/tokenmark-mcp
- Repository: https://github.com/sypherin/tokenmark-mcp
- Website: https://tokenmark.app/
- Changelog RSS feed: https://verifymcp.io/servers/sypherin-tokenmark-mcp/altronis-tokenmark-mcp.xml
- Changelog JSON feed: https://verifymcp.io/servers/sypherin-tokenmark-mcp/altronis-tokenmark-mcp.json
- HTML version of this page: https://verifymcp.io/servers/sypherin-tokenmark-mcp/altronis-tokenmark-mcp