# Picodev Lab (remote · lab.picodev.fr)

Arduino benches simulated in the browser: sessions, sketch push, simulation, tests, challenges.

- Trust score: 63/100 (medium)
- Registry status: active
- Liveness: live
- Owner verified: no
- Last scored: 2026-10-22

## Components

- remote · `lab.picodev.fr`: 63/100 (this document), [markdown](https://verifymcp.io/servers/picodev-sas-picodev-lab/lab.md), [page](https://verifymcp.io/servers/picodev-sas-picodev-lab/lab)

## Channel facts

- Endpoint: `https://lab.picodev.fr/mcp`
- Transports: `streamable-http`
- Auth: `none`
- Version: `0.9.1`

## Trust breakdown

How this component scores in each security and reliability category. Every signal is checked automatically against the live server, 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 2026-10-22.

- **Endpoint Security**: 66/100
  - The endpoint's TLS certificate is valid, in date, and uses a strong key.
  - Authorisation check failed: no authorisation is required to call this server, and it exposes a tool marked destructive (close_session).
  - HTTPS is enforced; there's no plaintext access path.
  - The HSTS (Strict-Transport-Security) header is present.
  - DNSSEC is configured correctly; the domain's records validate against the full chain to the root.
- **Transport & Reachability**: 100/100
  - Verified streamable-http transport via a live MCP handshake.
- **Schema Quality & AI Usability**: 56/100
  - AI-judged instruction clarity (good).
  - Context-footprint check failed: tool/resource definitions use about 2975 tokens (~141/item across 21 items; 21 tools + 0 resources), over budget; trim descriptions and params.
  - Usage-examples check failed: none of the tools include examples.
- **Stability & Change Management**: 3/100
  - Stability observed for 1 of 30 days with no destabilising changes; credit accrues until the full window elapses.
- **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 (19% 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 3 tool(s) whose name or description implies an irreversible operation declare an MCP destructiveHint annotation.
  - An AI judge read all 22 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 Picodev Lab MCP server?

Picodev Lab is a hosted endpoint at https://lab.picodev.fr/mcp, so there is nothing to install locally. 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 --transport http picodev-sas-picodev-lab 'https://lab.picodev.fr/mcp'
```

### Cursor

```json
{
  "mcpServers": {
    "picodev-sas-picodev-lab": {
      "url": "https://lab.picodev.fr/mcp"
    }
  }
}
```

### VS Code

```json
{
  "servers": {
    "picodev-sas-picodev-lab": {
      "type": "http",
      "url": "https://lab.picodev.fr/mcp"
    }
  }
}
```

### Codex

```toml
[mcp_servers.picodev-sas-picodev-lab]
url = "https://lab.picodev.fr/mcp"
```

### opencode

```json
{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "picodev-sas-picodev-lab": {
      "type": "remote",
      "url": "https://lab.picodev.fr/mcp",
      "enabled": true
    }
  }
}
```

### OpenClaw

```bash
openclaw mcp add picodev-sas-picodev-lab --url 'https://lab.picodev.fr/mcp' --transport streamable-http
```

### Hermes

```yaml
mcp_servers:
  picodev-sas-picodev-lab:
    url: "https://lab.picodev.fr/mcp"
```

### Netclaw

```json
{
  "McpServers": {
    "picodev-sas-picodev-lab": {
      "Transport": "http",
      "Url": "https://lab.picodev.fr/mcp"
    }
  }
}
```

### Vellum

```bash
assistant mcp add picodev-sas-picodev-lab -t streamable-http -u 'https://lab.picodev.fr/mcp'
```

### Other

```json
{
  "mcpServers": {
    "picodev-sas-picodev-lab": {
      "type": "http",
      "url": "https://lab.picodev.fr/mcp"
    }
  }
}
```

The mcpServers block is a cross-client convention. Remote transports vary, so check your client's docs.

## 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.

### 2026-10-22 (score 63, +1)

- [functional improvement] Stability: unverified → 0.03

### 2026-10-21 (score 62)

First indexed and scored.

## MCP tools (21)

### `list_setups` (~24 tokens)

List setups

The available setups, their modes (free, challenge) and their benches.

Output parameters:

- `result` (array)

### `open_session` (~50 tokens)

Open a session

Create a session (mode: free or challenge). Give open_url to the human: nothing simulates until they open it in their browser.

Input parameters:

- `mode` (string, required)
- `setup` (string, required)

### `push_to_editor` (~92 tokens)

Push sources to the editor

Replace the code in the human's browser and, if run, compile it in the lab's cloud and start it. The fallback
    of push_program: use it only when the toolchain to compile locally is missing (no arduino-cli, core or library).

Input parameters:

- `agent_token` (string, required)
- `files` (array, required)
- `run` (boolean)
- `session` (string, required)

Output parameters:

- `result` (string)

### `push_program` (~152 tokens)

Push a compiled program

Send a program compiled on the visitor's machine (Intel HEX text, 48 KiB at most, e.g. arduino-cli's
    <sketch>.ino.hex) to the human's browser, which loads it in the simulator without compiling and, if run, starts it.
    files: the sources, shown read-only in the editor, never compiled. During a turn on a real bench, the bench gets it
    too. Prefer it to push_to_editor, which is the fallback when the local toolchain is missing.

Input parameters:

- `agent_token` (string, required)
- `files`
- `hex` (string, required)
- `run` (boolean)
- `session` (string, required)

Output parameters:

- `result` (string)

### `read_simulation` (~122 tokens)

Read the simulation

Last state (with circuit, the browser's components and wires), last compilation and the events after since_ms
    (2000 at most). In challenge mode: the level's brief and its results. circuit_modified: the circuit differs from
    the setup sheet's (types, ids, properties, wires; not positions or colours), null when unknown. circuit_error: the
    editor's last refusal of a pushed circuit, returned once.

Input parameters:

- `agent_token` (string, required)
- `session` (string, required)
- `since_ms` (number)

### `set_visibility` (~50 tokens)

Set visibility

Change the visibility (private, link, public). Returns the viewer link; current viewers are disconnected.

Input parameters:

- `agent_token` (string, required)
- `session` (string, required)
- `visibility` (string, required)

### `queue_join` (~54 tokens)

Join the bench queue

Take a place in the queue for benches of this kind (for example real or sim). The browser must be connected.

Input parameters:

- `agent_token` (string, required)
- `kind` (string, required)
- `session` (string, required)

### `queue_status` (~55 tokens)

Queue status

The session's place: state none, waiting (position, estimated wait eta_s) or turn (bench, end of the turn ends_at, Unix seconds).

Input parameters:

- `agent_token` (string, required)
- `session` (string, required)

### `queue_leave` (~39 tokens)

Leave the bench queue

Give back the place in the queue, or the turn if it has started.

Input parameters:

- `agent_token` (string, required)
- `session` (string, required)

### `close_session` (~50 tokens)

Close the session

End the session now, as when it expires: the place or turn ends, viewers and the browser are disconnected, everything is forgotten.

Input parameters:

- `agent_token` (string, required)
- `session` (string, required)

### `session_status` (~113 tokens)

Session status

The session's state without its events, cheap enough to poll: browser connected (owner), visibility, current viewer link (not renewed), viewers, created_at and expires_at (Unix seconds), run (compilations so far: event times restart at each), level and levels (null in free mode), capture ({id, kind, state} of the session's capture, logic or analog, or null). Not reported to the human's journal.

Input parameters:

- `agent_token` (string, required)
- `session` (string, required)

### `next_level` (~93 tokens)

Next level

Challenge mode: move to the next level, as the bench's "Next level" button, and only when it would be offered:
    every test of the current level passed and a next level exists. Returns {level, levels}, the new level (index);
    session_status's level changes with the bench's next state. No level skipping.

Input parameters:

- `agent_token` (string, required)
- `session` (string, required)

### `capture_logic` (~119 tokens)

Capture logic signals

Record every edge of the requested pins (probes of the setup sheet; by default the logic ones) for duration_ms
    (1 to 10 000 ms). In the browser's simulation, in simulated time; during the session's turn on a real bench, at the
    bench's logic analyser, in real time (no browser or running simulation needed then). Read the result with read_capture.

Input parameters:

- `agent_token` (string, required)
- `duration_ms` (integer)
- `pins`
- `session` (string, required)

### `capture_analog` (~177 tokens)

Capture analog signals

Sample the voltage of 1 to 4 board pins (0-13 digital, 14-21 = A0-A7; by default the setup's scope probes) at
    rate_hz (100 to 10 000) for duration_ms (1 to 10 000 simulated ms), 200 000 samples at most in all. An ADC input
    reads the voltage the circuit applies; any other pin, its logic level times the supply (PWM shows as a square wave).
    Simulation only: refused during a turn on a real bench, whose analyser has no analog channel. Read the result with
    read_capture.

Input parameters:

- `agent_token` (string, required)
- `duration_ms` (integer)
- `pins`
- `rate_hz` (integer)
- `session` (string, required)

### `read_capture` (~255 tokens)

Read a capture

State of a capture (recording, ready, failed) and its kind (logic, analog). Once ready, the measures per pin
    and, for a logic capture, the edges [t_us, pin, v], 2000 from offset; for an analog one, rate_hz and the samples,
    rows [mv_1, ...] in the order of pins (row k at k / rate_hz), 2000 rows from offset. next: the following offset, or
    null. t0_ms: the first sample on the session's clock (ms since the simulation's launch, or since the start of the
    turn on a real bench), null when unknown; t_us and rows count from the capture's start. format="png": the same
    without edges or samples, plus a chart of the signals as an image (from_ms and to_ms, from the capture's start,
    to zoom; 4 per minute).

Input parameters:

- `agent_token` (string, required)
- `capture` (string, required)
- `format` (string)
- `from_ms`
- `offset` (integer)
- `session` (string, required)
- `to_ms`

### `read_setup` (~66 tokens)

Read a setup

A served setup sheet (setup.json, files included; never a solution): board, components and wires (circuit),
    probes, levels. Plus, when it exists, schema.png, the starting circuit as the editor draws it, as an image.

Input parameters:

- `id` (string, required)

### `push_circuit` (~151 tokens)

Push a circuit

Free mode only: replace the circuit in the human's browser (the simulation stops). components:
    [{type, id, properties, x?, y?}] (the board is not a component), wires: [{id, start: {componentId, pinName},
    end: {...}, color?}], in the shape of the setup sheet's circuit; 40 components, 120 wires, 48 KiB. The
    editor checks types and pins: read the outcome with read_simulation (circuit, circuit_error). 6 per minute.

Input parameters:

- `agent_token` (string, required)
- `components` (array, required)
- `session` (string, required)
- `wires` (array, required)

Output parameters:

- `result` (string)

### `bench_snapshot` (~73 tokens)

Bench photo

During the session's turn on a real bench: a photo of the bench, as an image, and t_ms, the instant of the
    photo in ms since the start of the turn (null when unknown). One every 2 seconds.

Input parameters:

- `agent_token` (string, required)
- `session` (string, required)

### `spec_draft` (~362 tokens)

Draft a project spec

Draft the specification of the visitor's own project with Picodev's interview: you ask the visitor, this tool says what to ask next and makes the PDF. Nothing is sent and nothing is kept. Call it with the sections written so far (none at first). sections: {id: text in simple Markdown}, ids envelope, fonctionPrincipale, fonctions, alimentation, format, environnement, production, preferences, 4000 characters at most each. sections.contact, optional: the contact details the visitor wants printed at the end of the document (1000 characters at most; not part of the interview's progress). language: fr or en, the one the visitor answers in (headings, look_for and questions, the PDF). pdf=true: also the PDF, with Picodev Lab's template; requires title (one line). figures, with pdf=true, optional: the drawings the visitor approved, 3 at most, each {section (a section id), caption (one line), type (png or svg), data (base 64, 200 KiB at most decoded)}, placed after their section's text; an SVG loads nothing (no script, no DTD, href # or data:image/ only). Returns {rules (how to run the interview: follow them), sections [{id, heading, required, state (empty, thin, filled), look_for and questions while not filled}], next (the section to ask about, or null), ready (the four required sections filled: enough for Picodev to start), filename and pdf (base 64) with pdf=true}.

Input parameters:

- `figures`
- `language` (string)
- `pdf` (boolean)
- `sections`
- `title`

### `brief_submit` (~475 tokens)

Send the spec to Picodev

Send Picodev the specification of the visitor's project, as a PDF, or only return its preview (send=false). sections: {id: text in simple Markdown}, ids envelope, fonctionPrincipale, fonctions, alimentation, format, environnement, production, preferences (the 8 sections of a product specification), 4000 characters at most each; a missing section reads "À préciser" (to be specified). send=false: preview, nothing is sent, contact and consent are not required. send=true: requires contact {name, email, company?, phone?} and consent=true. nda=true: the visitor wants a non-disclosure agreement first; without sections, only contact and title are sent, without PDF. session and agent_token, optional: the mail also says what the visitor did in the lab. language: fr or en, the PDF's headings. sections.contact and figures: as for spec_draft (refused with nda=true, as sections). Write the sections with spec_draft first. Before passing consent=true, show the visitor the consent text in the visitor's language, and wait until they accept it. In French: "J'accepte que mes coordonnées et la description de mon projet soient envoyées à Picodev par courriel, pour qu'ils me répondent. Picodev les garde 3 ans après le dernier échange et ne les utilise pour rien d'autre. Détails : lab.picodev.fr/confidentialite" In English: "I agree that my contact details and the description of my project are sent to Picodev by e-mail, so that they can answer me. Picodev keeps them for 3 years after our last exchange and uses them for nothing else. Details: lab.picodev.fr/en/privacy" Returns {sent, filename, pdf (base 64, absent with nda), missing (empty or too short sections), next (a sentence for the visitor, to say in their language)}}: write the PDF on the visitor's machine.

Input parameters:

- `agent_token`
- `consent` (boolean)
- `contact`
- `figures`
- `language` (string)
- `nda` (boolean)
- `sections`
- `send` (boolean)
- `session`
- `title` (string, required)

### `request_meeting` (~334 tokens)

Request a meeting with Picodev

Send Picodev a meeting request. contact: {name, email, company?, phone?}. context: the project in a few sentences (2000 characters at most): what the visitor wants to do, where they stand, what they did in the lab. availabilities: 1 to 10 slots {from, to}, ISO 8601 dates with a time zone, within the next 60 days; or {text}, one sentence ("Tuesday mornings"). Requires consent=true. session and agent_token, optional: the mail also says what the visitor did in the lab. Before passing consent=true, show the visitor the consent text in the visitor's language, and wait until they accept it. In French: "J'accepte que mes coordonnées et la description de mon projet soient envoyées à Picodev par courriel, pour qu'ils me répondent. Picodev les garde 3 ans après le dernier échange et ne les utilise pour rien d'autre. Détails : lab.picodev.fr/confidentialite" In English: "I agree that my contact details and the description of my project are sent to Picodev by e-mail, so that they can answer me. Picodev keeps them for 3 years after our last exchange and uses them for nothing else. Details: lab.picodev.fr/en/privacy" Returns {sent, next}: Picodev answers by e-mail to confirm a slot.

Input parameters:

- `agent_token`
- `availabilities` (required)
- `consent` (boolean, required)
- `contact` (required)
- `context` (string, required)
- `session`

## Diagnostics

Captured diagnostic sections: TLS, DNSSEC, Authorisation, Transports. The full working is on the page: https://verifymcp.io/servers/picodev-sas-picodev-lab/lab#diagnostics

## Score history

- 2026-10-22: 63
- 2026-10-21: 62

## Common questions

### What is the Picodev Lab MCP server?

Picodev Lab is an MCP server listed in the public MCP registry as io.github.PICODEV-SAS/picodev-lab. Arduino benches simulated in the browser: sessions, sketch push, simulation, tests, challenges. This page covers its hosted endpoint (https://lab.picodev.fr/mcp).

### Is the Picodev Lab MCP server safe to use?

Picodev Lab scores 63 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 Picodev Lab MCP server expose?

Picodev Lab exposes 21 tools: list_setups, open_session, push_to_editor, push_program, read_simulation, and 16 more. Their descriptions and schemas cost roughly 2,906 tokens of context every time the server is loaded.

### Does the Picodev Lab MCP server require authentication?

No. We connected to Picodev Lab without credentials and it answered, so anything it exposes is reachable by anyone who knows the address.

### Is the Picodev Lab MCP server still maintained?

Picodev Lab is still listed as active in the MCP registry. We last reached this channel on 22 October 2026. Those dates come from our own scans of the registry and the channel itself, not from anything the publisher announced.

## Links

- Remote endpoint: https://lab.picodev.fr/mcp
- Repository: https://github.com/PICODEV-SAS/picodev-lab-plugin
- Website: https://lab.picodev.fr/
- Changelog RSS feed: https://verifymcp.io/servers/picodev-sas-picodev-lab/lab.xml
- Changelog JSON feed: https://verifymcp.io/servers/picodev-sas-picodev-lab/lab.json
- HTML version of this page: https://verifymcp.io/servers/picodev-sas-picodev-lab/lab