Spec Driven Development
MCP Server that facilitates spec-driven development workflows, not just Vibe Coding.
Descripción del creador (original en inglés)
Prepare -> execute for onchain agent actions: policy pre-flight, simulation, receipt log.
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.
Necesitas
¿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": {
"kit": {
"command": "npx",
"args": [
"-y",
"@satohub/kit",
"mcp"
]
}
}
}Ejecuta esto en tu terminal
claude mcp add kit -- npx -y @satohub/kit mcpPégalo en .cursor/mcp.json de tu proyecto, o en ~/.cursor/mcp.json
{
"mcpServers": {
"kit": {
"command": "npx",
"args": [
"-y",
"@satohub/kit",
"mcp"
]
}
}
}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": {
"kit": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"@satohub/kit",
"mcp"
]
}
}
}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.
Esta 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 6 nov 2026
El nombre en el registro se verificó con el dominio satohub.ai (el registro comprueba quién es dueño del dominio)
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 (0.1.0) llegó al registro hace 38 días.
Las apps compatibles con MCP, como Claude Desktop, Claude Code, Cursor y VS Code. También necesitas Node.js en tu computadora.
En la misma categoría: Spec Driven Development, Remote Desktop Commander, Whisper
Servidores de la misma categoría hechos por otros creadores, ordenados por estrellas.
Cómo leer una tarjeta:★ 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.
One routable IPv6 per AI agent, verifiable from the DNS root. Keyless tools need no signup.
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.
0.1.1 preview.
Sato Kit wraps the tools agents already use (viem, signers, x402) with a prepare -> execute contract: an action prepares an unsigned intent, a policy pre-flight explains any refusal, the transaction is simulated, and execute({ intent_id }) hands it to your signer and appends a hash-chained sato.receipt/v1 line.
What it is not: not a wallet, not a custodian, not an agent framework, and not a guarantee. It never holds funds and ships no keys.
network defaults to fork; mainnet needs an explicit opt-in.scan block of policy.json records payto_changed (caution | refuse | off), but it is enforced only by hosts that implement it. The offline guard in satohub-core (scanRecipient) does not enforce it today: a changed x402 payTo needs the payee's earlier value, which the offline check is not given.venue: "direct" skips Sato entirely.Contracts (sato.action/v1, sato.policy/v1, sato.receipt/v1) live in src/spec (see VENDORED.md).
npm install @satohub/kit viem
npx -y @satohub/kit@0.1 --help
Node 20 or later. viem is a peer; every framework SDK below is an optional peer, installed only if you use its subpath.
evaluatePreflight runs before an intent is returned and names each refused rule with its limit and the observed value. Once a key is in play the kit enforces nothing: enforcement is the signer's (a CDP, Privy or Turnkey policy compiled from the same policy.json, OWS's policy engine, or a Safe's owners). viemLocalSigner enforces only its mainnet refusal.x402.prepare reads what an x402 resource asks to be paid and returns a priced, policy-checked intent. execute does not send x402 payments: it sends EVM transactions, EIP-712 typed data (Safe proposals) and Solana transactions only. Pay through your x402 client.@CreateAction decorator posts an invocation event to cca-lite.coinbase.com on every invoke. Actions consumed through fromAgentKitProvider still make that call; the kit neither blocks nor adds to it.max_usd_* rule refuses with unknown_price rather than guessing.Generated from the registered descriptors by node scripts/gen-actions-doc.mjs (after npm run build); test/readme-actions.test.ts fails when this table drifts from the registry. "custody" is the descriptor's own answer to three sato.custody/v1 questions. toolset says whether the tool is in the default MCP profile or only under --toolsets all.
| action id | tool name | kind | effects | custody | chains | toolset |
|---|---|---|---|---|---|---|
chain.read | chain_read | read | read | reads key: no; key leaves: no; moves funds: never | ethereum, sepolia, base, base-sepolia, arbitrum, arbitrum-sepolia, optimism, optimism-sepolia, polygon | default, all |
swap.quote | swap_quote | read | quote | reads key: no; key leaves: no; moves funds: never | ethereum, base, arbitrum, optimism, polygon | default, all |
swap.prepare | swap_prepare | prepare | quote, sign, broadcast | reads key: no; key leaves: no; moves funds: with_approval | ethereum, base, arbitrum, optimism, polygon | default, all |
x402.prepare | x402_prepare | prepare | sign, pay | reads key: no; key leaves: no; moves funds: with_approval | base, base-sepolia, ethereum, sepolia, polygon, arbitrum, optimism, solana, solana-devnet | default, all |
erc8004.lookup | erc8004_lookup | read | read | reads key: no; key leaves: no; moves funds: never | ethereum, base, arbitrum, optimism, polygon | all |
tx.simulate | tx_simulate | read | simulate | reads key: no; key leaves: no; moves funds: never | ethereum, sepolia, base, base-sepolia, arbitrum, arbitrum-sepolia, optimism, optimism-sepolia, polygon | all |
token.approvals.list | token_approvals_list | read | read | reads key: no; key leaves: no; moves funds: never | ethereum, sepolia, base, base-sepolia, arbitrum, arbitrum-sepolia, optimism, optimism-sepolia, polygon | all |
token.approvals.revoke | token_approvals_revoke | prepare | sign, broadcast | reads key: no; key leaves: no; moves funds: with_approval | ethereum, sepolia, base, base-sepolia, arbitrum, arbitrum-sepolia, optimism, optimism-sepolia, polygon | all |
erc8004.register | erc8004_register | prepare | sign, broadcast | reads key: no; key leaves: no; moves funds: with_approval | ethereum, base, arbitrum, optimism, polygon | all |
bridge.quote | bridge_quote | read | quote | reads key: no; key leaves: no; moves funds: never | ethereum, base, arbitrum, optimism, polygon | all |
bridge.prepare | bridge_prepare | prepare | quote, sign, broadcast | reads key: no; key leaves: no; moves funds: with_approval | ethereum, base, arbitrum, optimism, polygon | all |
safe.info | safe_info | read | read | reads key: no; key leaves: no; moves funds: never | ethereum, sepolia, base, base-sepolia, arbitrum, optimism, polygon | all |
safe.propose | safe_propose | prepare | sign | reads key: no; key leaves: no; moves funds: with_approval | ethereum, sepolia, base, base-sepolia, arbitrum, optimism, polygon | all |
solana.read | solana_read | read | read | reads key: no; key leaves: no; moves funds: never | solana, solana-devnet | all |
solana.transfer | solana_transfer | prepare | sign, broadcast | reads key: no; key leaves: no; moves funds: with_approval | solana, solana-devnet | all |
solana.swap.quote | solana_swap_quote | read | quote | reads key: no; key leaves: no; moves funds: never | solana | all |
solana.swap.prepare | solana_swap_prepare | prepare | quote, sign, broadcast | reads key: no; key leaves: no; moves funds: with_approval | solana | all |
| tool name | kind | effects | toolset |
|---|---|---|---|
execute | execute | sign, broadcast | default, all |
status | meta | read | default, all |
actions_search | meta | read | default, all |
actions_describe | meta | read | default, all |
Default MCP profile, in order: chain_read, swap_quote, swap_prepare, x402_prepare, execute, status, actions_search, actions_describe.
Actions consumed from an AgentKit provider (fromAgentKitProvider) are not in this table: they are built at runtime from the provider you pass, and appear under toolset all.
| import | what it gives you |
|---|---|
@satohub/kit | createKit, coreActions, the tool surface, config loader, everything below re-exported |
@satohub/kit/spec | the vendored contracts: sato.action/v1, sato.policy/v1, sato.receipt/v1, id rules, descriptor lint |
@satohub/kit/policy | evaluatePreflight; compilers compileCdpPolicy, compilePrivyPolicy, compileTurnkeyPolicy |
@satohub/kit/signers | viemLocalSigner, owsSigner, cdpSigner, solanaLocalSigner, humanApprove |
@satohub/kit/intent | intent store, HMAC intent ids (loadOrCreateIntentSecret, verifyIntentId) |
@satohub/kit/receipts | the hash-chained sato.receipt/v1 log |
@satohub/kit/actions | every core action, descriptor and builder |
@satohub/kit/ai-sdk | satoKitTools(kit, opts) for the Vercel AI SDK |
@satohub/kit/agentkit | satoKitActionProvider(kit, opts) and fromAgentKitProvider(provider, { chain }) |
@satohub/kit/claude-agent-sdk | createSatoKitSdkServer(kit, opts) + requireApprovalHook() |
@satohub/kit/openai-agents | satoKitOpenAITools(kit, opts) |
@satohub/kit/langchain | satoKitLangChainTools(kit, opts) |
@satohub/kit/eliza | satoKitElizaPlugin(kit, opts) (compatibility adapter, see below) |
@satohub/kit/mcp | the local MCP server (stdio) |
@satohub/kit/cli | the sato-kit CLI as a function |
@satohub/kit/sato-os | attachToSatoOs, proposeIntent |
@satohub/kit/goat | fromGoatPlugin(plugin, { chain }) — consume an existing GOAT plugin with the kit's guarantees |
viemLocalSigner — a key in this process. Fork/testnet by default; refuses mainnet chain ids unless allowMainnet: true (not recommended). generate: true makes a throwaway in-memory key.owsSigner — a structural adapter over an Open Wallet Standard account; OWS's policy engine enforces. See src/signers/OWS.md.cdpSigner — a structural adapter over a Coinbase CDP server account; CDP enforces its policies. Pass a compiled policy as nativePolicy to carry it with the signer (the adapter does not register it).solanaLocalSigner — local Solana signing for devnet and tests only; refuses any non-devnet payload and a mainnet-beta RPC, with no mainnet switch.humanApprove(signer, approve) — wraps any signer so a person approves each summary before it signs.compileCdpPolicy, compilePrivyPolicy and compileTurnkeyPolicy turn one sato.policy/v1 into the provider's native policy document. Each result lists what did not compile (not_compiled) with the reason, so a rule the provider cannot express is never silently dropped.
sato-kit read <action> --input '<json>' run a read-only action
sato-kit prepare --input ' ' build an intent; nothing is signed sato-kit execute --intent hand one prepared intent to the signer sato-kit mcp [--toolsets default|all] local MCP server over stdio sato-kit doctor check config, Sato Status and template drift
Config comes from the working directory: policy.json (absent means the default policy, network fork), .sato/ (intents, intent secret, receipts), SATO_RPC_URL_ per chain, and SATO_SIGNER=viem-local-fork (fork network only) for a local signer.
npx -y @satohub/kit@0.1 mcp runs a stdio MCP server (registry name ai.satohub/kit). The default profile is the eight tools listed above; --toolsets all adds every other registered action. Tools that sign carry anthropic/requiresUserInteraction in _meta.
status and sato-kit doctor read the nightly per-action results (sato.action-status/v1) published by the templates repo. When that file cannot be read, the answer says so; nothing is filled in.
The file intent store and the intent HMAC secret live under .sato/ by default (intents/, intent.key, mode 0600). Keep them out of version control:
.sato/
License: MIT. The bundled Onchain Action Descriptor code (dist/spec) is Apache-2.0, copyright 2026 Prime Signal LLC: see LICENSE-ODA and NOTICE.
sato-kit doctor: template driftIn a repo made by create-sato-agent (one with sato.create.json), doctor also sends sato.create.json and sato.lock.json to https://satohub.ai/api/create/drift (3 s timeout) and prints the changes it returns: a newer template version that passed its checks, a custody change in a stored Sato Check profile for a pinned package, or a pinned package version that fails the templat
Solo se muestra el principio; lee el resto en el repo