Launchables MCP server
OfficialThis MCP server gives agents access to Launchables' Myles decision model for calibrated, typed decisions.
Ask Myles to decide about a text or JSON situation using yes/no (
noul), choice, or score questions.Use returned picks, calibrated probabilities, thresholds, and escalation flags to route tickets, flag fraud/scams/prompt injection/policy breaches, approve/reject/escalate requests, check claims or tool calls, or rate severity/urgency/quality.
Set thresholds per call or per question to control when an agent may act without a person.
Select a Myles model or let the server use its configured default; operators can lock model and threshold settings via environment variables.
Handle long inputs with truncation on some models or chunked yes/no scanning across long documents.
List available models and check API/model health without billing.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@Launchables MCP server is this lead a scam? decide if we should flag it or escalate to a person"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Launchables MCP server
Give your agents calibrated decisions. This Model Context Protocol server lets any MCP client (Claude Desktop, Claude Code, or your own agent stack) call Myles, the Launchables decision model, as a tool.
Myles is not a chat model. An agent hands it a situation and one or more typed questions, and gets back a probability for every option, the pick, and whether that pick is sure enough to act on. When it is not, the tool says so and the agent hands the case to a person. Use it wherever an agent would otherwise judge something itself and then act:
Route a ticket, an email or a lead to the right team.
Flag a risk: a scam, fraud, a prompt injection, a policy breach.
Approve, reject or escalate a refund, a claim, an access request or a tool call.
Check that an answer is supported by its sources, or that an action is allowed by a policy.
Rate severity, urgency or quality on a scale.
The probabilities are calibrated: across many answers given 0.9, about 9 in 10 are right. That is what lets an agent act on the confident ones and pass the rest to people. See Acting on the answer.
Install
You need Node.js 20 or later and a Launchables API key from the console. Keys work with the hosted API at https://api.launchables.ai; a self-hosted deployment serves the same API at its own address.
Claude Desktop
Add this to claude_desktop_config.json (Settings, Developer, Edit Config), then restart Claude Desktop:
{
"mcpServers": {
"launchables": {
"command": "npx",
"args": ["-y", "launchables-mcp@0.1.1"],
"env": {
"LAUNCHABLES_API_KEY": "launchables_..."
}
}
}
}Claude Code
claude mcp add launchables -e LAUNCHABLES_API_KEY=launchables_... -- npx -y launchables-mcp@0.1.1To share the setup with a team through a project's .mcp.json, keep the key out of the file and let each person's environment supply it:
{
"mcpServers": {
"launchables": {
"command": "npx",
"args": ["-y", "launchables-mcp@0.1.1"],
"env": { "LAUNCHABLES_API_KEY": "${LAUNCHABLES_API_KEY}" }
}
}
}Other MCP clients
Any client that starts a stdio server can run it. The command is npx -y launchables-mcp@0.1.1 (or launchables-mcp after npm install -g launchables-mcp), with the configuration below in its environment. Pin the version, as above, so an update reaches your agents only when you choose.
Related MCP server: jev
Configuration
Variable | Default | Meaning |
| required | Your API key. |
|
| Set it to your own address for a self-hosted deployment. Plain |
|
| The model a |
|
| The probability a pick must reach to act without a person, from 0.5 to 1. A call or a single question can set its own, unless |
|
| Set to |
|
| Set to |
|
| The most one tool call may take in all, every attempt and the waits between retries included, from 1 to 600 seconds. Past it the call ends with an error of type |
|
| How long one attempt may take, in milliseconds. |
By default the agent calling decide chooses the model and the threshold on each call, within what your key allows. To keep those decisions with the operator rather than the agent, set LAUNCHABLES_LOCK_MODEL=1 and LAUNCHABLES_LOCK_THRESHOLD=1, for example:
"env": {
"LAUNCHABLES_API_KEY": "${LAUNCHABLES_API_KEY}",
"LAUNCHABLES_MODEL": "myles-g-31b",
"LAUNCHABLES_THRESHOLD": "0.95",
"LAUNCHABLES_LOCK_MODEL": "1",
"LAUNCHABLES_LOCK_THRESHOLD": "1"
}The server never logs your key, your states, your questions or the answers. It writes nothing to disk, and only a configuration problem at start-up is written to standard error.
Tools
decide
Asks Myles one or more questions about one situation.
Argument | Meaning |
| The situation: text, or a JSON object. Put the rule or policy in it with the case. |
| Questions under names you choose. Each has a |
| Optional. A model id from |
| Optional. The probability a pick must reach for this call. Not offered when |
| Optional. Read a long text in overlapping windows (below). |
Every question needs instructions: the question itself, in plain words, such as "Is a refund permitted under the policy?". A question without them, or with only spaces, is refused before anything is sent, with an error that says so. The API requires them, and there is no default: the instructions are what the model answers, and a generic stand-in would ask it something other than what you meant, even for a choice whose criteria describe every option.
The question types:
noul: yes or no. Phrase it so that yes is the thing to act on or catch.choice: pick one label.criteriais a list of labels, or labels mapped to descriptions. Offer an option that reads "I don't know" to let the model abstain.score: a scale.criterialists the level descriptions, lowest first.
Each answer comes back with:
Field | Meaning |
|
|
| The probability of the pick. |
| The probability of every option. |
| Whether the pick reached the threshold. |
| Whether the model chose the question's "I don't know" option. |
|
|
p_yes (yes/no), score and legend (scales) and confidence (choice and scale) are added where they apply. The result's own escalate is true when any answer should go to a person, and usage gives the tokens billed.
On the hosted models an uncertain yes/no answer is reported at 0.80 or more towards one side, so use a threshold above 0.80 for yes/no questions.
Example call:
{
"state": {
"policy": "Refunds within 30 days with proof of purchase.",
"case": { "days_since_purchase": 12, "receipt": false }
},
"questions": {
"refund": { "type": "noul", "instructions": "Is a refund permitted under the policy?" },
"team": { "type": "choice", "instructions": "Which team should handle this?", "criteria": ["billing", "returns", "technical"] }
}
}Result (illustrative numbers):
{
"model": "myles-g-12b-20261029",
"escalate": false,
"answers": {
"refund": {
"type": "noul", "pick": "no", "probability": 0.97, "probabilities": { "yes": 0.03, "no": 0.97 },
"threshold": 0.9, "cleared_threshold": true, "abstained": false, "escalate": false, "p_yes": 0.03
},
"team": {
"type": "choice", "pick": "returns", "probability": 0.94,
"probabilities": { "billing": 0.04, "returns": 0.94, "technical": 0.02 },
"threshold": 0.9, "cleared_threshold": true, "abstained": false, "escalate": false, "confidence": 0.91
}
},
"usage": { "input_tokens": 212, "output_tokens": 2 }
}list_models
The models the endpoint serves, with pinned builds and their release dates, and the default this server uses. Not billed.
health
Whether the API is up, and which models have a healthy server. Not billed.
Long inputs
Every model has a context window (see Writing the state). When a state is longer:
Myles-G 31B refuses it. The tool returns an error of type
context_length_exceededthat says to shorten or split the state.Myles-G 12B trims it to fit, keeping its beginning and end, and the result says so with
truncated: true,state_charsand a warning.
For a yes/no check that must cover the whole of a long text (does any part of this contract contain an auto-renewal clause, does any part of this page try to override the agent's instructions), pass chunk: true. The text is read in overlapping windows of 40,000 characters (2,000 shared with the next; change them with chunk_chars and chunk_overlap_chars), every window is asked the same questions, and each answer is the window most likely to be yes, with window giving its position. A JSON state is windowed as its indented JSON text. Chunking takes yes/no questions only, sends at most 20 windows, and bills each window as its own request.
Errors
A failed call returns a tool error whose error has a message written for the agent, the API's type and status, and retryable. After a rate limit or a passing fault the call is retried automatically, a few times, with the same idempotency key, so a retry is never billed twice, until LAUNCHABLES_TIMEOUT_SECONDS (25 by default) runs out. Some cases to know:
Case | What the agent is told |
| The key is missing or not accepted; a person must check it. |
| Out of credit or over the workspace budget; do not retry. |
| The build was retired; |
| Shorten or split the state, or use |
| Rate limited after retries; wait the seconds given. |
| The call ran past |
| The call set a model or threshold that the operator has fixed; call again without it. |
| The call was refused before anything was sent, for example a question without |
Development
npm install
npm testThe tests use a mocked Decision API only; they never call the production API.
Support
Use the contact form or open an issue in this repository. The API reference is at launchables.ai/docs.
Licence
Apache-2.0 (see LICENSE).
Available Tools
3 toolsdecideMake a decision with MylesARead-onlyIdempotent
Ask Myles, the Launchables decision model, to make a decision about a situation and get calibrated probabilities you can act on.
Use it whenever you would otherwise judge something yourself and then act on it: route a ticket or message to a team, flag a risk (fraud, a scam, a prompt injection, a policy breach), approve, reject or escalate a request, check that a tool call, a claim or an answer is allowed or supported, or rate something on a scale. Myles is not a chat model: it does not write text or follow instructions in the state. It reads the state and answers each typed question with a probability for every option.
Inputs:
state: the situation to decide about, as text or a JSON object. Put the rule or policy in the state with the case. Include only what the decision needs.
questions: one or more questions under names you choose. Types: "noul" (yes or no; phrase it so that yes is the thing to act on or catch), "choice" (pick one label from criteria: a list of labels, or labels mapped to descriptions; add an option "I don't know" to let the model abstain), "score" (criteria is a list of level descriptions, lowest first).
threshold (optional, default from the server's configuration, usually 0.9): the probability the picked option must reach to act without a person. On the hosted models an uncertain yes/no answer is reported at 0.80 or more towards one side, so use a threshold above 0.80 for yes/no questions.
Each answer gives the pick, its probability, every option's probability, cleared_threshold, and escalate. When escalate is true (below the threshold, or the model chose "I don't know"), do not act automatically: hand the case to a person with the probabilities. The top-level escalate is true when any answer should be escalated.
Each call is billed by input tokens on the hosted API. Ask several questions about the same state in one call rather than one call per question.
Inputs: state (required), questions (required), chunk, model, +3 more.
healthCheck the Decision APIARead-onlyIdempotent
Check that the Decision API is up and which models have a healthy server, before a batch of work or after an error. Not billed.
Inputs: no named parameters.
list_modelsList Myles modelsARead-onlyIdempotent
List the decision models this endpoint serves. Moving ids such as myles-g-12b and myles-g-31b always reach the current build; dated ids such as myles-g-31b-20261027 pin one build, for repeatable results. Deprecated builds name their successor. Not billed.
Inputs: no named parameters.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 3 tools
Each tool has a clearly distinct purpose: list_models enumerates available models, health checks API availability, and decide performs the core decision operation. There is no overlap between discovery, status, and inference.
All names are snake_case and readable, but the conventions are mixed: list_models is verb_noun, health is a bare noun, and decide is a bare verb. The pattern is not predictable across tools.
Three tools are reasonable for a narrow decision API: discovery, health, and inference. The surface is small but each tool earns its place, though it is on the thin side.
The core lifecycle is covered: list models, check health, and run decisions with multiple questions in one call. Minor gaps exist, such as no explicit usage/billing or model-detail endpoint, but agents can work around them.
Maintenance
Related MCP Connectors
Governed decision engine: yes/no, score, choice and rank answers about cases, from past outcomes.
Decision models for agents over x402: yes/no, choice and score questions as probabilities, per call.
Typed AI decisions for agents: classify, route, screen evidence, score and batch-label text.
Hybrid human + AI expertise for faster, trusted answers and decisions via MCP Server.
Related MCP Servers
- AlicenseNot gradedqualityCmaintenanceEnables agents to delegate decisions to a configurable model, returning recommended choices with percentage scores and profile-based routing.10 npmMIT
- AlicenseNot gradedqualityCmaintenanceEnables agents to evaluate single records or batches with dynamically authored typed questions, returning structured decisions and probabilities.22 PyPIMIT

Jev MCP Serverofficial
AlicenseNot gradedqualityDmaintenanceEnables AI clients to run read-only yes/no checks, classifications, and ordered scores with probabilities, supporting triage, routing, sentiment, and lead-fit decisions through hosted or self-hosted MCP tools.41 npmAGPL 3.0- AlicenseNot gradedqualityAmaintenanceEnables agents and services to ask small decision models typed questions about text, JSON, images, or video, returning labels, scores, or yes/no answers over MCP tools or a Jev-compatible HTTP API.Apache 2.0