# dashAI (unofficial) (pypi · dashai-mcp)

Train, track and query models on a running dashAI instance. Read-oriented, no delete tools.

- Trust score: 65/100 (medium)
- Registry status: active
- Liveness: live
- Owner verified: no
- Last scored: 2026-09-20

## Components

- pypi · `dashai-mcp`: 65/100 (this document), [markdown](https://verifymcp.io/servers/maarmapa-dashai-mcp/dashai-mcp.md), [page](https://verifymcp.io/servers/maarmapa-dashai-mcp/dashai-mcp)

## Channel facts

- Registry: `pypi`
- Package: `dashai-mcp`
- Version: `0.3.1`
- 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 2026-09-20.

- **Supply Chain Security**: 100/100
  - No malware found by supply-chain analysis.
  - No known CVEs affecting this package version or its production dependencies.
  - Runs hatchling.build at install time, a recognised native-build step with no shell scripting around it.
  - 1 of 32 dependencies flagged as unhealthy.
- **Provenance & Transparency**: 32/100
  - Source repository is publicly reachable at the declared URL.
  - Provenance check failed: no build-provenance attestation is published.
  - License check failed: the license (MIT License) isn't a recognized OSI-approved license.
  - Actively maintained (last published 25 days ago).
  - Disclosure check failed: no security disclosure policy was found in the source repository.
- **Schema Quality & AI Usability**: 70/100
  - AI-judged instruction clarity (excellent).
  - Context-footprint check failed: tool/resource definitions use about 1764 tokens (~176/item across 10 items; 10 tools + 0 resources), over budget; trim descriptions and params.
  - Usage-examples check failed: none of the tools include examples.
- **Stability & Change Management**: 0/100
  - Stability not yet verified: not enough scan history yet (needs a 30-day window).
- **Tool Coverage**: 71/100
  - 100% of tools have a non-trivial description (not blank, and not just the tool's name).
  - 0% 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.
  - We read all 10 captured tool definition(s), and no name or description among them implies an irreversible operation.
  - An AI judge read all 10 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 dashAI (unofficial) MCP server?

dashAI (unofficial) runs locally as a PyPI package, launched with uvx dashai-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 maarmapa-dashai-mcp -- uvx dashai-mcp
```

### Cursor

```json
{
  "mcpServers": {
    "maarmapa-dashai-mcp": {
      "command": "uvx",
      "args": [
        "dashai-mcp"
      ]
    }
  }
}
```

### VS Code

```json
{
  "servers": {
    "maarmapa-dashai-mcp": {
      "command": "uvx",
      "args": [
        "dashai-mcp"
      ]
    }
  }
}
```

### Codex

```bash
codex mcp add maarmapa-dashai-mcp -- uvx dashai-mcp
```

### opencode

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

### OpenClaw

```bash
openclaw mcp add maarmapa-dashai-mcp --command uvx --arg dashai-mcp
```

### Hermes

```yaml
mcp_servers:
  maarmapa-dashai-mcp:
    command: "uvx"
    args: ["dashai-mcp"]
```

### Netclaw

```json
{
  "McpServers": {
    "maarmapa-dashai-mcp": {
      "Transport": "stdio",
      "Command": "uvx",
      "Arguments": [
        "dashai-mcp"
      ]
    }
  }
}
```

### Vellum

```bash
assistant mcp add maarmapa-dashai-mcp -t stdio -c uvx -a dashai-mcp
```

### Other

```json
{
  "mcpServers": {
    "maarmapa-dashai-mcp": {
      "command": "uvx",
      "args": [
        "dashai-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.

### 2026-09-16 (score 65, +15)

- [security improvement] Malware scan: unverified → pass

### 2026-09-15 (score 50)

First indexed and scored.

## MCP tools (10)

### `dashai_server_info` (~277 tokens)

Checks that dashAI is running and summarizes what is loaded.

Call this FIRST when something fails or when you do not know whether the
backend is up: it tells "dashAI is down" apart from "that id does not
exist", which are two problems with different fixes.

Args:
    params (NoArgs): no parameters.

Returns:
    str: JSON with the following schema:
    {
        "base_url": str,      # which instance is being targeted
        "reachable": bool,    # whether it responded
        "datasets": int,      # number of loaded datasets
        "runs": int,          # number of recorded runs
        "queue_empty": bool,  # whether the job queue is empty
        "compatibility": {    # live API vs the release verified end to end
            "verified_against": str,   # e.g. "dashAI 0.9.7.post1"
            "status": str,             # "ok" | "mismatch" | "unknown"
            "warnings": [str],         # only on mismatch: what differs
            "note": str                # only on mismatch/unknown
        }
    }
    On failure: "Error: <what happened and what to do>".

Input parameters:

- `params` (required)

Output parameters:

- `result` (string)

### `dashai_list_datasets` (~130 tokens)

Lists the datasets loaded in dashAI.

Returns only id, name, date and status — just enough to pick one. To see
columns and types use dashai_describe_dataset with the id.

Args:
    params (ListDatasets): contains:
        - limit (int): maximum to return, 1-200 (default 50)

Returns:
    str: JSON {"count": int, "datasets": [{"id", "name", "created", "status"}]}
    If there are none: a message explaining how to load data from the GUI.

Input parameters:

- `params` (required)

Output parameters:

- `result` (string)

### `dashai_describe_dataset` (~150 tokens)

Returns everything needed to configure a training run over a dataset.

Gathers into a single call what the raw API splits into four (`/{id}`,
\`/info`, `/types` and `/sample`), because deciding which columns are input
and which is output requires seeing them together.

Args:
    params (DescribeDataset): contains:
        - dataset_id (int): dataset id
        - include_sample (bool): include sample rows (default True)

Returns:
    str: JSON {"dataset": {...}, "info": {...}, "column_types": {...}, "sample": [...]}
    If one part is unavailable it comes back as null instead of failing whole.

Input parameters:

- `params` (required)

Output parameters:

- `result` (string)

### `dashai_list_components` (~146 tokens)

Lists the registered components: models, metrics, tasks and optimizers.

ALWAYS use this before dashai_train_model. The names dashAI expects are
exact and case-sensitive, and the catalogue changes with the plugins that
instance has installed — they cannot be guessed.

Args:
    params (ListComponents): contains:
        - types (Optional[List[str]]): filter by 'Model', 'Metric', 'Task', 'Optimizer'

Returns:
    str: JSON {"count": int, "components": [{"name": str, "type": str, "schema": {...}}]}
    The `schema` field describes the hyperparameters that component accepts.

Input parameters:

- `params` (required)

Output parameters:

- `result` (string)

### `dashai_train_model` (~327 tokens)

Trains a model on a dataset and returns the id of the enqueued job.

It does NOT wait for it to finish. Training can take minutes or hours, so
dashAI enqueues it and this tool returns immediately; progress is polled
with dashai_job_status.

Collapses the three calls the raw API demands:
  1\. POST /model-session/  creates the experiment (dataset, task, columns, metrics)
  2\. POST /run/            creates the run (model, hyperparameters)
  3\. POST /job/            enqueues the ModelJob

Args:
    params (TrainModel): contains:
        - dataset_id (int), task_name (str), model_name (str)
        - input_columns / output_columns (List[str])
        - metrics (List[str]), goal_metric (str)
        - parameters (Dict): model hyperparameters
        - splits (Dict[str, float]): proportions adding up to 1.0
        - optimizer_name (str), optimizer_parameters (Dict)
        - run_name (Optional[str])

Returns:
    str: JSON {"job_id": str, "run_id": int, "model_session_id": int, "status": "enqueued", "next_step": str}
    On failure: "Error: ..." stating which parameter dashAI rejected.

Examples:
    - "Train a random forest on dataset 3 predicting 'species'"
    - Do not use it to read results: that is dashai_get_run, with the run_id.

Input parameters:

- `params` (required)

Output parameters:

- `result` (string)

### `dashai_job_status` (~145 tokens)

Polls the status of an enqueued job (training, prediction, explanation).

dashAI's statuses: `not_started` (queued), `started` (running), `finished`
(done) and `error` (failed). Telling `started` from `error` matters: the
first is worth waiting on, the second does not improve by polling again.

Args:
    params (JobStatus): contains:
        - job_id (str): id returned when enqueuing

Returns:
    str: JSON {"job_id": str, "status": str, "finished": bool, "failed": bool, "raw": {...}}

Input parameters:

- `params` (required)

Output parameters:

- `result` (string)

### `dashai_list_runs` (~110 tokens)

Lists the recorded training runs, with their status.

Useful for comparing models trained within the same experiment.

Args:
    params (ListRuns): contains:
        - model_session_id (Optional[int]): filter by experiment
        - limit (int): maximum to return, 1-200 (default 50)

Returns:
    str: JSON {"count": int, "runs": [{"id", "name", "model_name", "status", "goal_metric"}]}

Input parameters:

- `params` (required)

Output parameters:

- `result` (string)

### `dashai_get_run` (~110 tokens)

Returns the configuration and metrics of a training run.

This is where results are read once dashai_job_status says `finished`. If
the run did not finish, the metrics will come back empty — that is not an
error.

Args:
    params (GetRun): contains:
        - run_id (int): run id

Returns:
    str: JSON with the full run: model parameters, status and metrics per
    split (train / validation / test).

Input parameters:

- `params` (required)

Output parameters:

- `result` (string)

### `dashai_predict` (~235 tokens)

Enqueues a prediction using the model of an already finished run.

Like training, it is asynchronous: it returns a job_id and the result is
followed with dashai_job_status.

Collapses the two calls the GUI makes (read from
DatasetPredictionPanel + createPrediction + enqueuePredictionJob, not
from the docs):
  1\. POST /predict/           creates the Prediction row ({run_id, dataset_id})
  2\. POST /job/  PredictJob   enqueues with {prediction_id} — NOT run_id

Passing only run_id to the job raises KeyError 'prediction_id' inside
PredictJob.run. The Prediction row must exist first.

Args:
    params (Predict): contains:
        - run_id (int): id of a finished run
        - dataset_id (Optional[int]): dataset to score; defaults to the
          run's training dataset (from its model session)

Returns:
    str: JSON {"job_id", "run_id", "prediction_id", "dataset_id",
    "status": "enqueued", "next_step"}

Input parameters:

- `params` (required)

Output parameters:

- `result` (string)

### `dashai_get_prediction` (~134 tokens)

Returns class counts for a finished prediction — never the rows.

dashAI stores predictions as an Arrow dataset on disk and GET /predict/
only returns the SQL row (id, status, paths). This tool reads that row
and, when the job is finished, aggregates the output column. The label
list never leaves the function.

Args:
    params (GetPrediction): contains prediction_id from dashai_predict.

Returns:
    str: JSON {prediction_id, run_id, dataset_id, status, finished,
    n, n_classes, class_counts}. Paths and row lists are stripped.

Input parameters:

- `params` (required)

Output parameters:

- `result` (string)

## Diagnostics

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

## Score history

- 2026-09-20: 65
- 2026-09-19: 65
- 2026-09-18: 65
- 2026-09-17: 65
- 2026-09-16: 65
- 2026-09-15: 50

## Common questions

### What is the dashAI (unofficial) MCP server?

dashAI (unofficial) is an MCP server listed in the public MCP registry as io.github.Maarmapa/dashai-mcp. Train, track and query models on a running dashAI instance. Read-oriented, no delete tools. This page covers its PyPI package (dashai-mcp).

### Is the dashAI (unofficial) MCP server safe to use?

dashAI (unofficial) scores 65 out of 100 on VerifyMCP. We found no known CVEs affecting it as of 20 September 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 dashAI (unofficial) MCP server expose?

dashAI (unofficial) exposes 10 tools: dashai_server_info, dashai_list_datasets, dashai_describe_dataset, dashai_list_components, dashai_train_model, and 5 more. Their descriptions and schemas cost roughly 1,764 tokens of context every time the server is loaded.

### Is the dashAI (unofficial) MCP server still maintained?

dashAI (unofficial) is still listed as active in the MCP registry. We last reached this channel on 20 September 2026. Those dates come from our own scans of the registry and the channel itself, not from anything the publisher announced.

## Links

- PyPI project: https://pypi.org/project/dashai-mcp/
- Socket report: https://socket.dev/pypi/package/dashai-mcp
- Repository: https://github.com/Maarmapa/dashai-mcp
- Changelog RSS feed: https://verifymcp.io/servers/maarmapa-dashai-mcp/dashai-mcp.xml
- Changelog JSON feed: https://verifymcp.io/servers/maarmapa-dashai-mcp/dashai-mcp.json
- HTML version of this page: https://verifymcp.io/servers/maarmapa-dashai-mcp/dashai-mcp