nc-tools-mcp
This server does not currently have a deployable release on Glama. Follow the repository's setup instructions to install and configure it for your MCP client, including any required credentials and local dependencies.
nc-tools — the Typed Machine API for coding agents
A terminal-free machine interface: every operation the terminal used to provide (files, search, patches, git, processes, tests, packages, network) is a typed tool with structured inputs, structured outputs, structured errors, and a full append-only event journal. One static Rust binary speaks MCP over stdio.
Thesis: the terminal is the machine's untyped API. Replace it with a typed one and agents become verifiable, replayable, measurable, and safer — without changing their model or harness.
Quick start
Requires Rust (stable) and Node ≥ 20 (for the test/dev harnesses only — the kernel itself has zero runtime dependencies).
git clone https://github.com/ncdevshiv/nc-tools.git
cd nc-tools
cargo build --release -p nct-mcp # → target/release/nc-tools-mcp(.exe)
npm run conform # 27 protocol conformance cases, black-boxMount it in any MCP client (ZCode, Claude Desktop, Cursor-style hosts, custom
loops) — see docs/MCP-CONFIG.md. Point the command at
target/release/nc-tools-mcp with your workspace root as the only argument, or
run npm run install:bin to copy it to ~/.local/bin and use that stable path.
Related MCP server: Uten Agent Terminal MCP
What's in the box
Path | What it is |
| The Rust workspace — 8 crates; |
| Language-neutral protocol cases ( |
| Node test suite driving the binary as a black box over MCP stdio |
| Task suite, verifier validator, and agent-run harness (real verifiers, journaled transcripts) |
| Conformance runner, cross-audit, stub-audit, golden exporter, kernel client |
| Protocol spec, MCP client config, internet-tool design, audit trail |
The tool surface (60 typed tools)
fs.*— read/readMany/readRange, write/writeMany, append, copy, list, tree, stat, mkdir, delete, movepatch.apply/applyMany— search/replace edits with exactness guardssearch.grep/files/replace/semantic,code.symbols,text.diffgit.status/diff/add/blame/commit/log/branch/checkout/push/pullproc.spawn/start/status/readOutput/stop/list/kill/runScript/watch— argv-typed, no shelltest.run(node, pytest),pkg.add/list/scripts/runScriptnet.http/probePort/fetch/robots/search— see docs/INTERNET-TOOLS.mdenv.get/set/list,sys.snapshot/rollback/listSnapshots/journal/workspace/doctor,batch.execute
The frozen machine-readable export is conformance/golden/tools.json
(regenerate with npm run golden; changes must be deliberate). The full
contract — error taxonomy, journal schema, transport bindings — is
docs/PROTOCOL.md.
Principles
Typed in, typed out. Tools return structured results and structured errors with machine-readable remediation hints. No stdout scraping.
Everything is journaled. Every tool call and result is an event in
<workspace>/.nc-tools/journal.jsonl. Sessions are replayable and auditable.Mutations are deliberate. Writes go through validated patches (search/replace blocks with occurrence counts). No silent overwrites.
No shell. There is no
exec/bash tool.proc.spawnis argv-typed. If a task requires a shell, that is a tool to design, not an escape hatch.
Verification gates
Every claim in this repo is machine-checked:
cargo test # Rust unit + integration suites
npm run test:node # black-box protocol/behavior suite over MCP stdio
npm run conform # language-neutral conformance (27 cases)
npm run crossaudit # audits THIS repo through its own MCP server
npm run audit # stub/TODO/mock/placeholder scan (exit 1 on findings)
npm run verify:verifiers # bench verifiers: untouched→fail, solved→pass, wrong→rejectedLicense
MIT
This server cannot be deployed
Maintenance
Related MCP Connectors
Your org's AI agents, tasks, runs, search, and brain files as MCP tools and resources.
214 tools for AI agents over one MCP endpoint: image, video, audio, 3D, data, storage, mail, GPUs.
Governed app access for AI agents: 1,000+ apps & 12,000+ tools via Code Mode MCP.
MCP server for your apps' tools and custom tools, plus hosted AI agents and approval-gated workflows
Related MCP Servers
- AlicenseNot gradedqualityCmaintenanceProvides coding agents with structured planning, persistent project memory, automated verification, and safety permission controls through MCP tools, enabling better planning, context retention, self-checking, and guarded execution.13 npmMIT
- AlicenseBqualityCmaintenanceEnables AI agents to run shell commands, execute Python code, read and write files, search text, and create or delete folders through MCP stdio tools.9MIT
- AlicenseNot gradedqualityAmaintenanceEnables coding agents to inspect files, search text, compare directories, read symbols, and apply structured edits against large codebases through bounded, local-first MCP tools.4Apache 2.0
- AlicenseBqualityAmaintenanceEnables AI coding agents and local LLMs to store and recall project memory, compact verbose logs, look up code symbols, create and rewind workspace snapshots, detect risky commands and loops, pack context optimally, and draft tokens through MCP tools.102Apache 2.0