kryos-mcp
Governed code execution for AI agents that type-checks, runs, and capability-verifies Kryos code, ensuring agents only run code with declared and allowed capabilities.
README
kryos-mcp
Governed code execution for AI agents, backed by the Kryos compiler's capability system. An MCP server that lets any client type-check, run, and capability-verify Kryos code — so an agent can only run code whose reach (network, filesystem, process, crypto, …) is declared and checked at compile time.
Why
Most "run this code" tools give an agent an unconstrained interpreter. kryos-mcp runs code through Kryos, where every function declares its capabilities (@capabilities(net, io)) and the compiler rejects anything that reaches further. verify_capabilities turns that into a governance verdict: ALLOW only if the code type-checks, every declared capability is inside the grant, and every function is annotated. Because Kryos enforcement is opt-in per function, unannotated functions are reported as UNCONSTRAINED rather than waved through.
Tools
| Tool | What it does |
|---|---|
kryos_check(code, files?) |
Type-check without producing artifacts; surfaces type + capability errors. |
kryos_run(code, files?, timeout_ms?) |
Compile and run with a hard timeout; captures stdout/stderr/exit code. |
capability_manifest(code, files?, strict?) |
Per-function capability manifest (JSON). |
verify_capabilities(code, files?, granted[], allow_unannotated?) |
ALLOW/DENY against a capability grant; reports unconstrained functions. |
kryos_audit(code, files?) |
Full audit: capability usage, extern/FFI surface, secret-pattern scan. |
files is an optional map of sibling .kry module files (name → content) so use <mod> resolves.
Install
Requires the Kryos toolchain on PATH (or set KRYOS_BIN).
npm install
claude mcp add kryos --scope user -- node /abs/path/to/kryos-mcp/server.mjs
Verify
node smoke-test.mjs # spawns the server, exercises every tool incl. a DENY and an ALLOW verdict
Capabilities reference
Valid grants (case-insensitive): net, io, ffi, compute, crypto, process, env, term, db, time, all. Notable gotchas the manifest reflects: env_get/exit require process; time_now requires time; network builtins require net.
MIT.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
Neon Database
MCP server for interacting with Neon Management API and databases
E2B
Using MCP to run code via e2b.
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.