Spec Driven Development
MCP Server that facilitates spec-driven development workflows, not just Vibe Coding.
Descripción del creador (original en inglés)
MCP bridge to any SiLA 2 server (lab-automation standard).
Estos tres nunca se combinan en una sola puntuación: las opiniones, las mediciones y la popularidad son cosas distintas.
Todavía no escribimos un resumen sencillo para este. Abajo están los datos que leímos del registro.
Elige la app de IA que usas y copia la configuración.
LABMCP_ADDRESS, LABMCP_SIMULATE, LABMCP_READ_ONLY, LABMCP_LIMITSNecesitas
¿Qué app usas?
Elige tu app arriba. La recordamos, así que la próxima vez el botón dirá “Instalar en …”.
Abre Claude Desktop → Settings → Developer → Edit Config y pega esto en claude_desktop_config.json
{
"mcpServers": {
"labmcp-sila2": {
"command": "uvx",
"args": [
"labmcp-sila2"
],
"env": {
"LABMCP_ADDRESS": "your-value",
"LABMCP_SIMULATE": "your-value",
"LABMCP_READ_ONLY": "your-value",
"LABMCP_LIMITS": "your-value"
}
}
}
}Ejecuta esto en tu terminal
claude mcp add labmcp-sila2 --env LABMCP_ADDRESS=your-value --env LABMCP_SIMULATE=your-value --env LABMCP_READ_ONLY=your-value --env LABMCP_LIMITS=your-value -- uvx labmcp-sila2Completa los valores vacíos antes de ejecutarlo.
Pégalo en .cursor/mcp.json de tu proyecto, o en ~/.cursor/mcp.json
{
"mcpServers": {
"labmcp-sila2": {
"command": "uvx",
"args": [
"labmcp-sila2"
],
"env": {
"LABMCP_ADDRESS": "your-value",
"LABMCP_SIMULATE": "your-value",
"LABMCP_READ_ONLY": "your-value",
"LABMCP_LIMITS": "your-value"
}
}
}
}Tu navegador te pide permiso para abrir la app; luego la app muestra la configuración y te pide confirmar. Los valores de las claves quedan vacíos para que los completes tú.
Pégalo en .vscode/mcp.json de tu proyecto
{
"servers": {
"labmcp-sila2": {
"type": "stdio",
"command": "uvx",
"args": [
"labmcp-sila2"
],
"env": {
"LABMCP_ADDRESS": "your-value",
"LABMCP_SIMULATE": "your-value",
"LABMCP_READ_ONLY": "your-value",
"LABMCP_LIMITS": "your-value"
}
}
}
}Tu navegador te pide permiso para abrir la app; luego la app muestra la configuración y te pide confirmar. Los valores de las claves quedan vacíos para que los completes tú.
Este se ejecuta en tu computadora, y ChatGPT solo puede conectarse a servidores en internet. Elige otra app.
Tu elección se guarda en una cookie en este dispositivo; no necesitas cuenta.
LABMCP_ADDRESS envLABMCP_SIMULATE envLABMCP_READ_ONLY envLABMCP_LIMITS envEsta configuración se genera a partir de la ficha del creador en el registro. Si el README dice otra cosa, sigue el README.
Se ejecuta en tu computadora; la lista de herramientas aparece en tu app de IA después de instalarlo.
Revisamos información que cualquiera puede ver; no hemos leído el código. Las revisiones que no aplican no se cuentan. · Revisado el 9 nov 2026
El nombre en el registro se verificó con la cuenta de GitHub K-Dense-AI, dueña del repo
Todavía no hemos leído este repo en GitHub
Todavía no hemos leído este repo en GitHub
Este se ejecuta en tu computadora; no hay endpoint en internet
Se ejecuta en tu computadora y usa ajustes locales en su lugar
Se ejecuta en tu computadora; solo podemos leer listas de herramientas de endpoints en internet
Comparado con todos los servidores activos de esta categoría en el registro. Las estrellas solo cuentan cuando quien lo publica es dueño del repo.
Superó 1 de 1 revisiones básicas aplicables. No hemos leído el código. Antes de agregar una clave o un token, revisa qué acceso pide el servidor y usa una clave que tenga solo los permisos que necesitas.
El registro no indica ninguna clave obligatoria, aunque algunos servidores te piden iniciar sesión la primera vez que los usas.
Se ejecuta en tu computadora. Tu app de IA lo inicia cuando hace falta, y los datos no pasan por los servidores del creador (a menos que el propio servidor llame a un servicio en línea).
El repo de GitHub cambió por última vez –, y la última versión (0.1.2) llegó al registro hace 43 días.
Las apps compatibles con MCP, como Claude Desktop, Claude Code, Cursor y VS Code. También necesitas Python (uv) en tu computadora.
En la misma categoría: Spec Driven Development, Remote Desktop Commander, SwarmMemo
Servidores de la misma categoría hechos por otros creadores, ordenados por estrellas.
Cómo leer una tarjeta:↓ Descargas por semanai★ Estrellas de GitHubiTres pilares: reseñas · revisiones · usoiActualizado = fecha en que se publicó la última versión en el registroi
MCP Server that facilitates spec-driven development workflows, not just Vibe Coding.
Remote Desktop Commander: use files, terminal and processes on your own computers from any AI app.
Talk with agent swarms: rooms, private messages and paid tasks for AI agents
Copiado tal cual del repo del creador. PopNx MCP no lo ha revisado ni lo respalda. Se quitaron HTML, imágenes y scripts. Los enlaces llevan fuera de este sitio.
Let an AI agent work with any SiLA 2 server: liquid handlers, incubators, plate readers, robotic arms, balances or scheduling middleware that implement the open SiLA 2 lab-automation standard. The bridge reads the server's own Feature Definitions (FDL), so it knows every command, property, parameter type, unit and constraint. It validates parameters against the FDL before sending, runs and tracks long-running (observable) commands, and cancels them through SiLA's CancelController.
| Package | labmcp-sila2 |
| Instruments | Any SiLA 2 server: devices and middleware with SiLA 2 interfaces, and servers built with sila_python, sila_java, sila_csharp or sila_cpp |
| Interfaces | Ethernet: gRPC over HTTP/2 with TLS (or insecure test servers), mDNS server discovery |
| Protocol | SiLA 2 v1.x (Part A & Part B specifications, FDL schemas & core features, client library sila2 (docs)) |
| Status | 🧪 simulated: tested against a real in-process SiLA 2 server (sila2 library, gRPC on localhost), not yet verified against vendor servers. Report a hardware test |
uvx labmcp-sila2 --simulate --check
--simulate starts a real SiLA 2 server in-process, built with the sila2 library. It runs on 127.0.0.1 on a free port, unencrypted, with discovery off. It implements a simulated thermoblock feature (org.labmcp/simulation/TemperatureController/v1) and the official CancelController feature:
CurrentTemperature property, plus TargetTemperature, RampRate and DeviceStateControlTemperature, an observable command with progress, intermediate responses and a DeviceBusy defined errorRunProgram, taking a list of {TargetTemperature, HoldTime} structuresSetRampRate, constrained to 0 < rate ≤ 10 K/sSwitchOffThe whole bridge runs through the real gRPC and FDL path.
discover_servers tool (mDNS _sila._tcp), or run the server with its address and port shown. SiLA has no default port, but 50052 is common.--option root_cert=ca.pem: many servers write it to a file or publish it in their discovery record. Servers started in insecure test mode need --option insecure=true. For mutual TLS, add --option client_cert=… --option client_key=…. uvx labmcp-sila2 --address 192.168.1.40:50052 --option root_cert=~/sila/ca.pem --check
uvx labmcp-sila2 --address sila://localhost:50052 --option insecure=true --check # test server uvx labmcp-sila2 --option server_name="Liquid Handler 1" --option root_cert=ca.pem --check # via discovery
Claude Code
claude mcp add sila -- uvx labmcp-sila2 --address 192.168.1.40:50052 --option root_cert=/path/ca.pem
Claude Desktop / Cursor / Windsurf (claude_desktop_config.json, .cursor/mcp.json, …)
{
"mcpServers": { "sila": { "command": "uvx", "args": ["labmcp-sila2", "--address", "192.168.1.40:50052", "--option", "root_cert=/path/ca.pem", "--option", "command_allowlist=Incubator.*,PlateReader.ReadPlate"] } } }
--read-only lets the agent browse features, read and subscribe to properties, and check command status, but not call commands. cancel_command stays available. To restrict which commands may run, use --option command_allowlist=Feature.Command,Feature.*. Entries can also be separated by ; or spaces, which is handy inside LABMCP_OPTIONS=command_allowlist=A.B;C.*.
| Tool | Kind | Description |
|---|---|---|
call_command | ⚠️ hazard | Run a SiLA command. This may physically act on the device (move, dispense, heat, shake, open doors): tell the user what will happen first. Parameters are validated against the Feature Definition before anything is sent, and the command must be in the allow-list if one is set. Unobservable commands return their responses; observable ones return an execution_id. |
cancel_command | 🛑 safety | Cancel a running observable command, or all commands, through SiLA's CancelController feature. Always available (even read-only). If the server has no CancelController, this explains that and points to the device's own stop command. |
discover_servers | 👁 read | Find SiLA 2 servers on the local network with SiLA Server Discovery (mDNS _sila._tcp). Lists name, UUID, address and port of each without connecting. Does not need a configured server. |
get_command_log | 👁 read | Return the most recent raw commands sent to / replies received from the instrument (newest last). Useful for debugging and for recording what was done. |
get_command_result | 👁 read | Responses of a finished observable command, or the SiLA execution error it finished with. Fails with a clear message if the command is still running. |
get_command_status | 👁 read | Status of an observable command started with call_command: waiting / running / finishedSuccessfully / finishedWithError, progress, estimated remaining time and the latest intermediate response. |
get_connection_info | 👁 read | Report which instrument is connected (identity, address, simulated or real), whether the server is read-only, and the active safety limits. Call this first. |
get_property | 👁 read | Read the current value of a SiLA property (for observable properties: the first value of a subscription). |
get_server_info | 👁 read | Identify the connected SiLA server (name, type, UUID, version, vendor, description), list its features, and say whether commands can be cancelled and which commands this bridge may call. |
list_executions | 👁 read | All observable command executions started in this session, newest last, with their status. |
list_features | 👁 read | Describe the server's features from their Feature Definitions: every command (observable or not, parameters, responses, intermediate responses, defined errors, and whether the allow-list permits calling it) and every property, with types rendered as JSON-schema-like dicts including units and constraints. |
reconnect | 🛑 safety | Close and re-open the connection to the instrument (e.g. after it was power cycled or a cable was re-plugged). |
subscribe_property | 👁 read | Collect a property's values for duration_s seconds (SiLA subscription for observable properties, polling otherwise) and summarise them (e.g. watch a temperature settle). |
| Limit | Default | Meaning |
|---|---|---|
max_command_wait_s | 300 s | Longest time call_command may block waiting for an observable command |
max_subscription_duration_s | 120 s | Longest subscribe_property collection |
max_discovery_s | 30 s | Longest mDNS discovery scan |
Override at launch, e.g. --limit max_command_wait_s=60. The main guards for commands:
| Guard | How |
|---|---|
| Hazard annotation | call_command is marked destructive, so MCP clients ask before running it |
| Allow-list | --option command_allowlist=Feature.Command,...: other commands are refused before anything is sent, and list_features shows callable: false for them |
| FDL validation | parameter names must match exactly. Types, ranges (min/max, inclusive or exclusive), sets, lengths, patterns, element counts, fully qualified identifiers and structure fields are checked against the server's FDL before sending |
| Read-only mode | --read-only hides call_command |
| Cancellation | cancel_command (SAFETY) uses CancelController.CancelCommand / CancelAll when the server implements that feature, and otherwise says so and points to the device's own stop command |
| Deadlines | unobservable commands and property reads time out after call_timeout_s (default 30 s, at most 300 s) instead of hanging; a subscription opened for a timed-out read is cancelled, and an observable command the server accepts after the deadline still appears in list_executions so it can be cancelled |
The server always re-validates too, so parameters rejected there come back as SiLA validation errors.
list_features renders each SiLA type as a JSON-schema-like dict: Real becomes number, Integer becomes integer, List becomes array and Structure becomes object. SiLA units appear as x-unit (the label) and x-unit-si (the conversion to SI base units, *SI = value × factor + offset*). Other SiLA details use x-sila-* keys. Values are passed as JSON:Binary: base64 stringsDate: YYYY-MM-DD[±HH:MM]Time: HH:MM:SS[±HH:MM]Timestamp: ISO 8601, UTC if no offset is givenAny: {"type": "Real", "value": 1.5}, basic types onlyexecution_id right away. get_command_status reports the SiLA execution status (waiting, running, finishedSuccessfully, finishedWithError), progress, estimated remaining time and the latest intermediate response. get_command_result returns the responses, or the defined/undefined execution error together with the error's description from the FDL. Pass wait_s to call_command for short runs. Executions are tracked only for this session and are forgotten on reconnect.org.silastandard/core.commands/CancelController/v1, and get_server_info reports whether it does. Some vendors provide their own Stop/Abort commands instead, which list_features shows.LockController lock identifier or an AuthorizationService access token) can be passed with metadata={"LockController.LockIdentifier": "…"} on call_command, get_property and subscribe_property. Values are validated against the metadata's FDL type.DOCTYPE/entity declaration (SiLA FDL never uses one), so a hostile server can't trigger XML entity expansion. Features that the sila2 library rejects as invalid are left out of list_features (the command log records why).Schema and ContentType constraints, and the Unit constraint, which is informational. The server enforces these. Large binaries (over 2 MB) use SiLA binary transfer, which the sila2 library handles.sila2, which is why the simulator's RunProgram uses a plain list and checks the 1–10 steps itself. Responses of that shape are affected in this bridge: the server executes the command, but its response can't be decoded. The bridge reports this as "WAS EXECUTED … could not be decoded" so the command is not repeated blindly. Properties of that shape decode fSolo se muestra el principio; lee el resto en el repo