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., "@cint MCP Server explain diagnostic C1001 and how to fix it"
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.
cint-mcp)
Official public Model Context Protocol (MCP) server for cint 0.2.0 (checked against compiler identity d6ebe30d604346afe4a7912c5d6f3ce2b9d401641d361772892b44b9292609ff).
Gives coding agents (Claude Code, Claude Desktop, Cursor, Windsurf, Copilot, etc.) instant in-memory access to:
Public Specifications (SPEC-00P through SPEC-09P)
Diagnostic & Fault Catalogue (C1001 through C9102, E_OVERFLOW, E_ASSERT, etc.)
Verified .ci Recipes with cryptographic receipts
cint University Curriculum (CU-000.2 through CU-401.2)
Native Toolchain Execution (cint_check, cint_eval, cint_test) when a local cint binary is present
Zero external package dependencies — runs on standard Python 3.8+ on Linux, macOS, and Windows.
You can add this MCP server directly via Git or by cloning the repository:
uvx)claude mcp add cint -- uvx --from git+https://github.com/pennyleans/cint-mcp cint-mcp.mcp.jsonAdd to your project's .mcp.json:
{
"mcpServers": {
"cint": {
"command": "python",
"args": ["path/to/cint-mcp/cint_mcp_server.py"]
}
}
}claude mcp add cint -- python /path/to/cint-mcp/cint_mcp_server.pyAdd to your claude_desktop_config.json:
Windows: %APPDATA%\Claude\claude_desktop_config.json
macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
{
"mcpServers": {
"cint": {
"command": "python",
"args": ["F:\\cint-mcp\\cint_mcp_server.py"]
}
}
}Add to .cursor/mcp.json or windsurf config:
{
"mcpServers": {
"cint": {
"command": "python",
"args": ["cint_mcp_server.py"]
}
}
}Related MCP server: codebase-chat-mcp
Tool | Parameters | Description |
|
| Search across specifications, glossary, errata, and university lessons |
|
| Retrieve a self-contained specification or lesson chunk |
|
| Explain compile error codes ( |
|
| Fetch verified 0.2.0 |
|
| List curriculum lessons across |
|
| Compile |
|
| Compile and run a program ( |
|
| Run the program's |
The three execution tools need a local cint: on PATH, named by CINT_PATH, or the newest build under CINT_BUILD/boot/<leg>. CINT_MCP_TIMEOUT sets their time limit in seconds (default 120).
Resource URI | Description |
| Public edition SHA-256 hashes and compiler source identity |
| Quickstart guide for C programmers |
| Quickstart guide for Python and NumPy developers |
| Condensed cint 0.2.0 syntax, types, and rules |
| Refused constructs catalogue under SPEC-09P.1 |
Licensed under the Apache License, Version 2.0 with LLVM Exceptions matching the cint project. See LICENSE for details.
Compile cint (.ci) source without running it (cint emit-c) and return its CINT-DIAG-1 diagnostics, or 'compiles'. Needs a local cint (PATH or CINT_PATH); look codes up with cint_lookup_diagnostic.
Inputs: source (required).
Compile and run a cint (.ci) program (cint run) and return its output, or its CINT-DIAG-1 diagnostic or CINT-FAULT-1 fault record. A program is void main() { ... } or top-level statements. Needs a local cint (PATH or CINT_PATH).
Inputs: source (required).
Retrieve one self-contained specification or curriculum chunk by the ID cint_search_docs returns (e.g. 'spec-04p:scope', 'cu-201-04:muldiv').
Inputs: id (required), verbose.
Fetch a verified cint 0.2.0 code recipe (fixed-point-trig, fixed-point-ln-sqrt, wide-i1024-scaling, svg-generator, fnv1a-digest, money-ledger, kernel-double-buffer) with digest.
Inputs: name (required), verbose.
List available cint University lessons (CU-000.2 to CU-401.2).
Inputs: course, verbose.
Explain any compile diagnostic (C1001 to C9102) or runtime fault (E_OVERFLOW to E_ASSERT) with before/after fixes.
Inputs: code (required), verbose.
Search across cint 0.2.0 specifications (SPEC-00P to SPEC-09P), GLOSSARY, ERRATA, and University curriculum (CU-000 to CU-401). Compact output by default.
Inputs: query (required), limit, scope, verbose.
Run the test "name" { ... } blocks of cint (.ci) source (cint test) and return one CINT-TEST-1 line per test, with the fault record of each failure. Needs a local cint (PATH or CINT_PATH).
Inputs: source (required).
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Scored across 8 tools
Each tool targets a distinct action on a distinct resource: check (compile only), eval (compile+run), test (run test blocks), lookup_diagnostic (explain codes), search_docs vs get_doc (find vs retrieve), get_recipe, list_curriculum. The descriptions explicitly disambiguate the easily confused compile/run/test trio and the search/get doc pair.
All tools share the cint_ prefix and snake_case, and most follow verb_noun (lookup_diagnostic, get_recipe, list_curriculum, search_docs, get_doc). The bare verbs check, eval, and test deviate slightly from the pattern but remain readable and unambiguous.
Eight tools is well-scoped for a language-tooling server: execution (check/eval/test), documentation discovery (search_docs/get_doc), diagnostics explanation, and curated assets (recipes/curriculum). Each tool earns its place with no filler.
The surface covers the full core workflow of compiling, running, testing, diagnosing, and looking up docs/recipes/lessons. Minor gaps exist (no version/environment info tool, no formatter or project scaffolding), but agents can work around these.
AI-native programming language and toolchain for accountable, verifiable agent-driven development.
Project memory for coding agents: requirements, decisions, code graph and delivery telemetry.
Project memory, documentation and controlled secret access for AI coding agents.
Your team's shipping standards, org map and delivery metrics, inside your coding agent.
