# Local AI MCP (npm · @tmhs/local-ai-mcp)

Run, fit, and route work to local LLMs (Ollama, LM Studio, llama.cpp, MLX, vLLM, and more).

- Trust score: 80/100 (high trust)
- Registry status: active
- Liveness: live
- Owner verified: no
- Last scored: 2026-11-05

## Components

- npm · `@tmhs/local-ai-mcp`: 80/100 (this document), [markdown](https://verifymcp.io/servers/tmhsdigital-local-ai-mcp/tmhs-local-ai-mcp.md), [page](https://verifymcp.io/servers/tmhsdigital-local-ai-mcp/tmhs-local-ai-mcp)

## Channel facts

- Registry: `npm`
- Package: `@tmhs/local-ai-mcp`
- Version: `0.6.1`
- 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 2026-11-05.

- **Supply Chain Security**: 98/100
  - No malware found by supply-chain analysis.
  - No known CVEs affecting this package version or its production dependencies.
  - No install/post-install scripts declared.
  - 30 of 92 dependencies flagged as unhealthy.
- **Provenance & Transparency**: 87/100
  - Source repository is publicly reachable at the declared URL.
  - Cryptographically verified build provenance (signed, bound to TMHSDigital/local-ai-mcp).
  - License check failed: the license (CC-BY-NC-ND-4.0) isn't a recognized OSI-approved license.
  - Actively maintained (last published 1 days ago).
  - Publishes a security disclosure policy (SECURITY.md).
- **Schema Quality & AI Usability**: 80/100
  - 100% of prompts and resources have a non-trivial description (not blank, and not just the item's name).
  - AI-judged instruction clarity (good).
  - Context-footprint check failed: tool/resource definitions use about 3109 tokens (~124/item across 25 items; 19 tools + 6 resources), over budget; trim descriptions and params.
  - Usage-examples check failed: none of the tools include examples.
- **Stability & Change Management**: 7/100
  - Stability observed for 2 of 30 days with no destabilising changes; credit accrues until the full window elapses.
- **Tool Coverage**: 98/100
  - 100% of tools have a non-trivial description (not blank, and not just the tool's name).
  - 93% of tool parameters carry a description.
  - Structured output schemas are declared (100% of tools); any adoption earns full credit.
- **Tool Safety**: 100/100
  - No prompt-injection markers were found in the server instructions, tool names or descriptions we captured.
  - All 1 tool(s) whose name or description implies an irreversible operation declare an MCP destructiveHint annotation.
  - An AI judge read all 21 captured unit(s) of tool text and found none that tries to manipulate the model reading it.
- **Capabilities**: 100/100
  - Implements a supported MCP spec version (2025-11-25); the latest is 2026-07-28.

## Install

### How do I install the Local AI MCP server?

Local AI MCP runs locally as an npm package, launched with npx -y @tmhs/local-ai-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 tmhsdigital-local-ai-mcp -- npx -y @tmhs/local-ai-mcp
```

### Cursor

```json
{
  "mcpServers": {
    "tmhsdigital-local-ai-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "@tmhs/local-ai-mcp"
      ]
    }
  }
}
```

### VS Code

```json
{
  "servers": {
    "tmhsdigital-local-ai-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "@tmhs/local-ai-mcp"
      ]
    }
  }
}
```

### Codex

```bash
codex mcp add tmhsdigital-local-ai-mcp -- npx -y @tmhs/local-ai-mcp
```

### opencode

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

### OpenClaw

```bash
openclaw mcp add tmhsdigital-local-ai-mcp --command npx --arg -y --arg @tmhs/local-ai-mcp
```

### Hermes

```yaml
mcp_servers:
  tmhsdigital-local-ai-mcp:
    command: "npx"
    args: ["-y", "@tmhs/local-ai-mcp"]
```

### Netclaw

```json
{
  "McpServers": {
    "tmhsdigital-local-ai-mcp": {
      "Transport": "stdio",
      "Command": "npx",
      "Arguments": [
        "-y",
        "@tmhs/local-ai-mcp"
      ]
    }
  }
}
```

### Vellum

```bash
assistant mcp add tmhsdigital-local-ai-mcp -t stdio -c npx -a -y @tmhs/local-ai-mcp
```

### Other

```json
{
  "mcpServers": {
    "tmhsdigital-local-ai-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "@tmhs/local-ai-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.

### 2026-11-05 (score 80, +1)

No change was recorded against any check on this day. Stability & Change Management went from 3 to 7. That category is still filling its 30-day observation window: 1 days of observed history at the previous scan, 2 at this one. The score rises as the window fills, whether or not the server changes.

### 2026-11-04 (score 79, +15)

- [security improvement] Malware scan: unverified → pass
- [functional improvement] Stability: unverified → 0.03

### 2026-11-03 (score 64)

First indexed and scored.

## MCP tools (19)

### `list_providers` (~100 tokens)

List providers

List configured model runtimes (Ollama, LM Studio, llama.cpp, Docker Model Runner, and any opt-in MLX / vLLM / Foundry Local / OpenAI-compatible / hosted instances) with host, kind (local or hosted), live status, and capabilities.

Input parameters:

- `provider` (string): Provider instance name. Omit to let the server choose (local providers only, unless LOCAL_AI_ALLOW_HOSTED_FALLBACK is set).

Output parameters:

- `providers` (array)

### `list_models` (~63 tokens)

List installed models

List models installed/available on each live provider, keyed by provider. Pass provider to query just one.

Input parameters:

- `provider` (string): Provider instance name. Omit to let the server choose (local providers only, unless LOCAL_AI_ALLOW_HOSTED_FALLBACK is set).

Output parameters:

- `models` (object)

### `list_loaded` (~66 tokens)

List loaded models

List models currently loaded into memory on each live provider, with VRAM use and context length where the runtime reports them.

Input parameters:

- `provider` (string): Provider instance name. Omit to let the server choose (local providers only, unless LOCAL_AI_ALLOW_HOSTED_FALLBACK is set).

Output parameters:

- `loaded` (object)

### `model_info` (~90 tokens)

Model info

Detailed metadata for one model (family, parameter size, quantization, context length, and layer architecture where available). Without provider, queries the live local providers that have the model.

Input parameters:

- `model` (string, required): Model id/name to inspect
- `provider` (string): Provider instance name. Omit to let the server choose (local providers only, unless LOCAL_AI_ALLOW_HOSTED_FALLBACK is set).

Output parameters:

- `info` (object)
- `model` (string)

### `pull_model` (~129 tokens)

Pull model

HEAVY: Download a model onto ONE provider (multiple GB, can take a long time). Streams download progress via MCP progress notifications when the client sends a progressToken. Without provider, uses the provider that already has the model, else the first live local provider that can pull.

Input parameters:

- `model` (string, required): Model id/name to pull (e.g. qwen3:8b, hf.co/<repo>:<quant>)
- `provider` (string): Provider instance name. Omit to let the server choose (local providers only, unless LOCAL_AI_ALLOW_HOSTED_FALLBACK is set).

Output parameters:

- `pull` (object)

### `remove_model` (~121 tokens)

Remove model

DESTRUCTIVE: Permanently delete a model from one provider. provider is REQUIRED (no fan-out). If the client supports elicitation the user is asked to confirm directly; otherwise confirm:true is required.

Input parameters:

- `confirm` (boolean): Must be true when the client cannot ask the user directly (no elicitation support)
- `model` (string, required): Model id/name to remove
- `provider` (string, required): Provider to delete from (required). Omit to let the server choose (local providers only, unless LOCAL_AI_ALLOW_HOSTED_FALLBACK is set).

Output parameters:

- `remove` (object)

### `load_model` (~148 tokens)

Load model

Load a model into memory on ONE provider so it is ready for inference. Optional keepAlive (e.g. '5m', '1h') and contextLength. Without provider, uses the live local provider that has the model.

Input parameters:

- `contextLength` (integer): Context window to load with (Ollama num_ctx, LM Studio --context-length)
- `keepAlive` (string): How long to keep the model resident, e.g. '5m'
- `model` (string, required): Model id/name to load
- `provider` (string): Provider instance name. Omit to let the server choose (local providers only, unless LOCAL_AI_ALLOW_HOSTED_FALLBACK is set).

Output parameters:

- `load` (object)

### `unload_model` (~83 tokens)

Unload model

Unload a model from memory to free VRAM/RAM on ONE provider. Without provider, uses the live local provider that has it loaded.

Input parameters:

- `model` (string, required): Model id/name to unload
- `provider` (string): Provider instance name. Omit to let the server choose (local providers only, unless LOCAL_AI_ALLOW_HOSTED_FALLBACK is set).

Output parameters:

- `unload` (object)

### `ensure_model` (~233 tokens)

Ensure model is ready

Make a model ready in one call: optionally pull it, check it fits current free VRAM, unload other models if needed (evict: 'lru' or 'all'), load it, and verify. Use dryRun:true to get the plan without changing anything.

Input parameters:

- `allowPull` (boolean): Download the model if it is not installed (default false)
- `contextLength` (integer): Context window to fit and load with (default 4096)
- `dryRun` (boolean): Return the plan only (default false)
- `evict` (string): Unload other models to make room: none (default), lru (least recently used first), all
- `keepAlive` (string): Keep-alive for the loaded model, e.g. '30m'
- `kvCacheType` (string): KV cache precision used for the fit estimate (default f16)
- `model` (string, required): Model id/name
- `provider` (string): Provider instance name. Omit to let the server choose (local providers only, unless LOCAL_AI_ALLOW_HOSTED_FALLBACK is set).

Output parameters:

- `fit` (object)
- `model` (string)
- `plan` (array)
- `provider` (string)
- `ready` (boolean)

### `health_check` (~77 tokens)

Health check

Check whether each provider is reachable and report its version (and live metrics such as vLLM KV-cache usage). Without provider, checks every configured provider, including hosted ones.

Input parameters:

- `provider` (string): Provider instance name. Omit to let the server choose (local providers only, unless LOCAL_AI_ALLOW_HOSTED_FALLBACK is set).

Output parameters:

- `health` (object)

### `system_resources` (~61 tokens)

System resources

Report local hardware: platform, total/available RAM, CPU count, and GPUs with total/free VRAM (NVIDIA via nvidia-smi, AMD/Intel via sysfs or the Windows registry, Apple Silicon unified memory). Each GPU names its data source.

Output parameters:

- `cpuCount` (number)
- `gpus` (array)
- `platform` (string)
- `ramFreeBytes` (number)
- `ramTotalBytes` (number)

### `fit_check` (~234 tokens)

Fit check

Will a model run on this machine, and where? Uses the installed model's size and exact layer architecture when the runtime exposes it (else parameter-count or catalog estimates) to size weights + KV cache for a context length, then compares with free VRAM summed across GPUs and available RAM. Returns target gpu / partial (GPU+CPU split) / cpu / none, with the numbers and how the KV estimate was made. Credits back memory the model already uses if it is loaded.

Input parameters:

- `contextLength` (integer): Context length for the KV-cache estimate (default 4096)
- `kvCacheType` (string): KV cache precision: f16 (default), q8_0, q4_0
- `model` (string, required): Model id/name to check
- `parameterSize` (string): Parameter size like "7B" (overrides lookup)
- `provider` (string): Provider instance name. Omit to let the server choose (local providers only, unless LOCAL_AI_ALLOW_HOSTED_FALLBACK is set).
- `sizeBytes` (number): Explicit weight size in bytes (overrides lookup)

Output parameters:

- `availableBytes` (number)
- `fits` (boolean)
- `loaded` (boolean)
- `model` (string)
- `note` (string)
- `requiredBytes` (number)
- `sizeSource` (string)
- `target` (string)

### `benchmark` (~216 tokens)

Benchmark

HEAVY: Runs REAL inference. Measures a model on this machine with a warmup (absorbs load time) plus N streamed runs: time-to-first-token, generation tokens/sec (runtime-reported when available, else measured after the first token), and prompt-processing speed. Pass models[] to compare several models with the same prompt, ranked by generation speed.

Input parameters:

- `maxTokens` (integer): Tokens to generate per run (default 128)
- `model` (string): Model id/name to benchmark
- `models` (array): Compare several models (ranked by median generation tokens/sec)
- `prompt` (string): Prompt (default: a short fixed prompt)
- `provider` (string): Provider instance name. Omit to let the server choose (local providers only, unless LOCAL_AI_ALLOW_HOSTED_FALLBACK is set).
- `runs` (integer): Measured runs per model (default 3)
- `warmup` (boolean): Run a 1-token warmup first so load time is excluded (default true)

Output parameters:

- `results` (array)

### `search_available` (~74 tokens)

Search catalog

Search the built-in curated catalog of well-known local models (Ollama tags) by name, family, or task. Offline and instant. For live results with per-quant fit, use search_models with source 'huggingface'.

Input parameters:

- `query` (string, required): Search text matched against model name, family, and tasks

Output parameters:

- `note` (string)
- `query` (string)
- `results` (array)

### `suggest_model` (~131 tokens)

Suggest model

Recommend catalog models for a task, ranked for THIS machine: fully-on-GPU first, then partial GPU offload, then CPU-only. Within a tier the largest (most capable) model comes first unless preferSpeed is true. Flags models already installed on a live provider and explains each pick.

Input parameters:

- `contextLength` (integer): Context length to plan for (default 4096)
- `kvCacheType` (string): KV cache precision (default f16)
- `preferSpeed` (boolean): Prefer smaller, faster models within each tier
- `task` (string, required): The task you want a model for

Output parameters:

- `suggestions` (array)
- `task` (string)

### `search_models` (~189 tokens)

Search models (live)

Find models to download, annotated with whether each fits this machine. source 'huggingface' searches GGUF repos on the Hugging Face Hub and returns per-quant sizes, the best quant for this hardware, and pullable references (Ollama: hf.co/<repo>:<quant>). source 'catalog' searches the offline curated list. The Ollama library has no public search API, so it is not a source.

Input parameters:

- `contextLength` (integer): Context length for fit (default 4096)
- `fitsOnly` (boolean): Only return results with a quant that fits
- `limit` (integer): Max repos (default 8)
- `query` (string, required): Search text, e.g. 'qwen coder' or 'gemma 3'
- `source` (string): Default huggingface
- `task` (string): Catalog source only: filter by task

Output parameters:

- `query` (string)
- `results` (array)
- `source` (string)

### `complete` (~386 tokens)

Delegated completion

DELEGATED INFERENCE: Offload a completion to a model you choose for cost savings and privacy. Without provider, runs on the live LOCAL provider that has the model (never a hosted API unless LOCAL_AI_ALLOW_HOSTED_FALLBACK is set); the result's 'provider', 'host' and 'local' fields say where it ran. Supports system prompts, JSON / JSON-Schema output (validated 'parsed' field), seed/topP/topK, image inputs for vision models, and contextLength for long inputs. Streams tokens via MCP progress notifications when the client sends a progressToken.

Input parameters:

- `contextLength` (integer): Context window to run with (Ollama num_ctx). Raise it for long inputs to avoid truncation.
- `maxTokens` (integer): Maximum tokens to generate
- `messages` (array): Chat messages; content may be a string or text/image parts (vision models)
- `model` (string, required): Model id/name to run the completion on
- `options` (object): Provider-specific extras passed through as-is
- `prompt` (string): Plain prompt text (alternative to messages)
- `provider` (string): Provider instance name. Omit to let the server choose (local providers only, unless LOCAL_AI_ALLOW_HOSTED_FALLBACK is set).
- `repeatPenalty` (number)
- `responseFormat`: Constrain output to JSON (json_object) or a JSON Schema (json_schema); the result includes a validated 'parsed' value
- `seed` (integer): Seed for reproducible sampling
- `stop` (array): Stop sequences
- `stream` (boolean): Stream tokens (default true). Progress notifications are sent when the client provides a progressToken.
- `system` (string): System instruction
- `temperature` (number): Sampling temperature
- `topK` (integer)
- `topP` (number)

Output parameters:

- `host` (string)
- `local` (boolean)
- `model` (string)
- `parsed`
- `provider` (string)
- `text` (string)

### `embed` (~104 tokens)

Delegated embeddings

DELEGATED EMBEDDINGS: Generate embeddings for a string or an array of texts. Without provider, runs on the live local provider that has the model. Always returns one vector per input.

Input parameters:

- `input` (required): Text or array of texts to embed
- `model` (string, required): Embedding model id/name
- `provider` (string): Provider instance name. Omit to let the server choose (local providers only, unless LOCAL_AI_ALLOW_HOSTED_FALLBACK is set).

Output parameters:

- `dimensions` (number)
- `embeddings` (array)
- `host` (string)
- `local` (boolean)
- `model` (string)
- `provider` (string)

### `delegate` (~457 tokens)

Delegate a task to the best local model

ONE CALL to offload work locally: pick the best model for a task on THIS machine, make it ready, and run it. Prefers models already loaded (no load cost), then installed models that fit your free VRAM, then (with allowPull) the best catalog model that fits. Task specialists (coder, vision, reasoning) win over general models; then larger fitting models win, grouped into size tiers (within ~20% parameter count); inside a tier, the model benchmarked fastest on this machine wins (run benchmark to record speeds; results expire after 14 days). Never uses a hosted provider unless allowHosted is true. Returns the completion plus a 'routing' block explaining the choice.

Input parameters:

- `allowHosted` (boolean): Allow hosted providers (default false)
- `allowPull` (boolean): Download a catalog model if nothing suitable is installed (default false)
- `contextLength` (integer): Context window to run with (Ollama num_ctx). Raise it for long inputs to avoid truncation.
- `maxSizeBytes` (number): Ignore models larger than this
- `maxTokens` (integer): Maximum tokens to generate
- `messages` (array): Chat messages; content may be a string or text/image parts (vision models)
- `options` (object): Provider-specific extras passed through as-is
- `prompt` (string): Plain prompt text (alternative to messages)
- `provider` (string): Restrict to one provider. Omit to let the server choose (local providers only, unless LOCAL_AI_ALLOW_HOSTED_FALLBACK is set).
- `repeatPenalty` (number)
- `responseFormat`: Constrain output to JSON (json_object) or a JSON Schema (json_schema); the result includes a validated 'parsed' value
- `seed` (integer): Seed for reproducible sampling
- `stop` (array): Stop sequences
- `stream` (boolean): Stream tokens (default true). Progress notifications are sent when the client provides a progressToken.
- `system` (string): System instruction
- `task` (string, required): What kind of work this is
- `temperature` (number): Sampling temperature
- `topK` (integer)
- `topP` (number)

Output parameters:

- `model` (string)
- `provider` (string)
- `routing` (object)
- `text` (string)

## Diagnostics

Captured diagnostic sections: Provenance, Dependencies. The full working is on the page: https://verifymcp.io/servers/tmhsdigital-local-ai-mcp/tmhs-local-ai-mcp#diagnostics

## Score history

- 2026-11-05: 80
- 2026-11-04: 79
- 2026-11-03: 64

## Common questions

### What is the Local AI MCP server?

Local AI MCP is listed in the public MCP registry as io.github.TMHSDigital/local-ai-mcp. Run, fit, and route work to local LLMs (Ollama, LM Studio, llama.cpp, MLX, vLLM, and more). This page covers its npm package (@tmhs/local-ai-mcp).

### Is the Local AI MCP server safe to use?

Local AI MCP scores 80 out of 100 on VerifyMCP. We found no known CVEs affecting it as of 5 November 2026. It declares no install or post-install scripts. Its build provenance is signed and verified. 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 Local AI MCP server expose?

Local AI MCP exposes 19 tools: list_providers, list_models, list_loaded, model_info, pull_model, and 14 more. Their descriptions and schemas cost roughly 2,962 tokens of context every time the server is loaded.

### Is the Local AI MCP server still maintained?

Local AI MCP 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 Local AI MCP server under?

Local AI MCP declares the CC-BY-NC-ND-4.0 licence, which is not on the OSI-approved list. Read the terms before using it at work, and note this covers the source only, not the cost of any service it calls.

## Links

- npm package: https://www.npmjs.com/package/@tmhs/local-ai-mcp
- Socket report: https://socket.dev/npm/package/@tmhs/local-ai-mcp
- Repository: https://github.com/TMHSDigital/local-ai-mcp
- Website: https://tmhsdigital.github.io/local-ai-mcp/
- Changelog RSS feed: https://verifymcp.io/servers/tmhsdigital-local-ai-mcp/tmhs-local-ai-mcp.xml
- Changelog JSON feed: https://verifymcp.io/servers/tmhsdigital-local-ai-mcp/tmhs-local-ai-mcp.json
- HTML version of this page: https://verifymcp.io/servers/tmhsdigital-local-ai-mcp/tmhs-local-ai-mcp