Glassbox Framework runs on your own machine — the client starts it, so there is no endpoint to ping. 28 installs a week from npm. Last commit 4 Sep 2026.
Runtime constitutional verification for AI answers — claim reasoning, ECS, red team, audits.
We read the source, 20 h ago · rules 3dff92dd89df
What this server is able to do. For an MCP server this is often the job itself — a terminal server runs commands because that is what it is for. Listed so you know what you are plugging in, not as an accusation.
этот файл ставится пользователю, но в репозитории его нет
const child = spawn(process.execPath, [tsNodeBin, indexTs], {
self._proc = subprocess.Popen(
m=__import__(f"nemoguardrails.library.{mod}.actions", fromlist=["actions"])
if (!url.startsWith("https://hooks.slack.com/")) throw new Error("Invalid Slack response URL.");
Is this your server and something here is wrong? Tell us — corrections are free and do not require a plan.
We found places where it runs commands, builds paths or queries from values it is given. None of that is a flaw by itself — it becomes one when the code changes, and code changes quietly between releases. We re-read it on every one.
This server runs on your own machine — install it with the package manager and the client starts it for you. Package name taken from the official registry entry.
claude mcp add glassbox-framework -- npx -y @glassbox-framework/mcp
{
"mcpServers": {
"glassbox-framework": {
"args": [
"-y",
"@glassbox-framework/mcp"
],
"command": "npx"
}
}
}
[mcp_servers.glassbox-framework]
command = "npx"
args = ["-y", "@glassbox-framework/mcp"]
{
"mcpServers": {
"glassbox-framework": {
"args": [
"-y",
"@glassbox-framework/mcp"
],
"command": "npx"
}
}
}
{
"mcpServers": {
"glassbox-framework": {
"args": [
"-y",
"@glassbox-framework/mcp"
],
"command": "npx"
}
}
}
This one needs environment variables set before it will start:
ANTHROPIC_API_KEY (Your Anthropic API key (sk-ant-...). Required for the claim extractor, red team, constitution engines, and ECS coherence checks. One v1 tool (generate_trust_card) works without an API key — it's pure assembly + deterministic audit hashing.), GLASSBOX_MODEL (Override the Claude model used by the verification engines. Defaults to claude-sonnet-4-6.), GLASSBOX_MAX_TOKENS (Per-engine-call max-tokens cap. Defaults to 2048; raise for verifying long-form content.), GLASSBOX_ECS_MODE (ECS aggregation mode: arithmetic (weighted mean, default) or geometric (stricter — any zero dimension collapses the total).).
The author declared them in the registry entry; get the values from the project itself.
Constitutional-AI safety guardrails for any LLM — personas, constitutions, adjudication.
Verify AI agent reasoning and logic for errors and contradictions. Pay-per-use.
Agent reasoning, memory, and token-optimized context for AI applications.
Execution-grounded verification for AI coding agents on Windows: evidence, not agent claims.
Independent effect verification and signed receipts for consequential AI agent actions.
Bitcoin research for AI agents: free answers, specialist Q&A, verification, evidence, timelines.
Runtime verification for agent tool calls: Ed25519 signed receipts, fail-closed RCE/SSRF checks.
Email verification for AI agents: send, hold or kill with the reason. Signup and credits by API.
Answers built from our own checks of this server.