SkillAgentSearch skills...

techne

An MCP server that gives AI agents the code navigation and refactorings of an IDE in ten languages, and states the evidence behind every answer

Install / Use

claude mcp add dokimasia -- npx -y github:dokimasia/techne

If the server publishes to npm under a different name, use that package instead — check the repo README.

About this skill
🔌

MCP Server

Model Context Protocol server

Quality Score

81/100

Supported Platforms

Claude Code
Claude Desktop

Our assessment of techne

techne scores 81/100 on our quality scale, 3474th of 4,606 Development & Engineering skills we index.

Its MCP Server is 6.7 KB long, well organised into 9 sections with 6 code examples: a thorough specification that gives an agent plenty to work with.

It has 3 GitHub stars, so there is little community track record yet; judge it on its content.

Substance
29/30
Structure
20/20
Description
15/15
Adoption
3/20
Freshness
15/15

Maintenance, license and trust

  • The repository was last updated today, so techne is actively maintained.
  • It is released under the MIT license, a permissive license that allows use, modification and commercial use with attribution.
  • Its trust signals score 92/100, with 1 caution from licensing, adoption, age or documentation. These come from repository metadata, not a code audit — read the skill file before letting an agent act on it.

Safety scan

No issues found

Our scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands.

Automated pattern scan on 2026-10-26. It catches known dangerous patterns, not every risk — read a skill before letting an agent act on it.

techne compared with similar skills

All 4 of these similar skills score higher than techne; compare them before choosing.

SkillScoreStarsUpdatedFormat
techne (this skill)by dokimasia813todayMCP Server
Agent-Reachby Panniantong100103.3k1d agoCLAUDE.md
taste-skillby Leonxlnx100102.9ktodayCLAUDE.md
headroomby headroomlabs-ai10077.9ktodayCLAUDE.md
ai-job-searchby ai-job-search-dev10066.3ktodayCLAUDE.md

Frequently asked questions

How do I install techne?
Run claude mcp add dokimasia -- npx -y github:dokimasia/techne. The install tabs above show the steps for each supported agent.
Which AI agents does techne work with?
It is written for Claude Code and Claude Desktop, as a MCP Server file. Other agents that read the same format can often use it too.
Is techne safe to use?
Our scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. It is MIT-licensed and scores 92/100 on trust signals. Skills are instructions an agent will follow, so read the file before installing it and do not approve commands you do not understand.
Is techne still maintained?
The repository was last updated today, so techne is actively maintained.
<!-- ~ Copyright Dokimasia B.V. 2026 ~ SPDX-License-Identifier: MIT -->

techne

techne is a server of the Model Context Protocol that gives an AI agent the code navigation and the refactorings of an IDE in ten languages: C, C#, Go, Java, JavaScript, Python, Ruby, Rust, Scala and TypeScript. One techne process serves a workspace over standard input and output. With --trust, it also serves each directory under a trusted folder that a call sets in its wd field.

Tools

| Tool | What it does | |---|---| | workspace | Lists the units of a directory, such as its packages, with their files, declarations and summaries | | outline | Lists the declarations of a file or a directory | | search | Finds declarations by name | | resolve | Returns the declaration that the name at a position denotes | | relations | Returns the relations of one kind of a declaration, such as its callers or its implementations | | verify | Runs the checks of a language over a file or a directory | | capabilities | Lists the engines of each language, with the tier and the cost of each role | | rename.symbol | Renames a declaration and every reference to it | | move.file | Moves a file and rewrites the references to it | | extract.function | Moves a run of lines into a new function and calls it | | document.symbol | Writes documentation onto a declaration | | apply.change | Writes the change of a preview |

A write tool previews its change unless the call sets dry_run to false. The preview contains the rewritten lines, the check that judged the result, and a handle that apply.change takes. techne writes a change in full or not at all.

Every answer states its evidence. The last line of the text states the fidelity and the completeness, such as resolved, total coverage. An empty answer adds an empty answer here means there are none when it proves that there are none. With --structured, the JSON of a result contains the same evidence:

  • fidelity is how the engine bound the names: syntactic, indexed or resolved.
  • completeness is how much of the scope the engine examined.
  • supportsNegativeClaim is true when the engine resolved the names and examined the whole scope. An empty answer proves that there are none only when it is true.

Language servers

The grammar of each language serves workspace, outline, search, relations and document.symbol without anything installed. The other tools need the language server of the language on PATH:

| Language | Program | |---|---| | C | clangd | | C# | csharp-ls | | Go | gopls, and the go command for the type checker that runs inside techne | | Java | jdtls | | JavaScript and TypeScript | typescript-language-server with the typescript package in the workspace, or tsc for a workspace on TypeScript 7 | | Python | pyright-langserver | | Ruby | ruby-lsp | | Rust | rust-analyzer | | Scala | metals |

A server reads the project files of its language, such as go.mod, Cargo.toml, build.sbt or compile_commands.json. capabilities lists each engine that cannot run, with the reason.

Build

techne uses cgo, because the grammars are C. You need Go 1.27 and a C compiler. From a checkout of this repository, run:

go build -o techne ./cmd/techne

Put the binary on your PATH.

Run

techne [-h | --help] [--version] [--structured] [--trust DIR]... [workspace]

techne serves the workspace at the path that you pass, or the working directory. While it serves, it writes only protocol messages to standard output, and its errors to standard error. -h and --help print the usage and exit. --version prints the version and exits.

--trust DIR adds a trusted folder, such as ~/Projects, and repeats for more than one folder. Every tool then takes the field wd, a directory under a trusted folder, absolute or relative to the workspace. A call with wd runs in the workspace of that directory, and the paths of the call and of its answer are relative to it. A call without wd runs in the workspace that techne serves. techne keeps three such directories open with their language servers. A call in a fourth directory closes the directory that served a call least recently and has no call in flight. The previews of a closed directory are gone. apply.change refuses their handles.

By default a result contains its answer as text written for a model, and techne declares no output schema. Claude Code passes the JSON of a result to the model when the result contains JSON, and the text is about half its size. With --structured, a result also contains the answer as JSON, and techne declares the output schema of each tool. Pass it when your client validates or processes the JSON.

A write takes an advisory lock of the operating system for the workspace, and the write of another techne process waits until the lock is free. Agents that each run their own techne over one repository do not interleave their writes.

Connect a client

In Claude Code, run:

claude mcp add techne -- techne /path/to/repository

A client that reads a JSON configuration starts techne with the same command and argument:

{
  "mcpServers": {
    "techne": {
      "command": "techne",
      "args": ["/path/to/repository"]
    }
  }
}

An editor that starts techne in the project it opens can pass the trusted folder alone, as "args": ["--trust", "~/Projects"]. A call then sets wd to any other project under that folder. techne expands the ~, because an editor passes it without a shell.

The Claude Code skill

skills/techne/SKILL.md is a skill for Claude Code. It covers:

  • the tool for each question about code
  • the evidence that ends each answer, and what an empty answer proves
  • the preview and the apply of a change

To install it for every project, link it into your skills directory from a checkout of this repository:

ln -s "$PWD/skills/techne" ~/.claude/skills/techne

A project can also include the skill in its own .claude/skills/ directory.

Development

The repository uses ergon for its checks and its releases. CONTRIBUTING.md states the setup.

make check        # the gate: lint, tests, race detector and vulnerability scan of every module
make test         # the tests of every module
make corpus       # drive the binary over large repositories, hours on a first run

make help lists every target. To drive the tools without a language server, set TECHNE_MOCK=1. techne then adds a mock language that serves every tool but extract.function at the resolved tier. docs/README.md lists the architecture documents, the RFCs, the decision records and the roadmap.

License

MIT. See LICENSE.

Related Skills

View on GitHub
GitHub Stars3
CategoryDevelopment
Updated11h ago
Forks0

Languages

Go

Trust signals

92/100

From repository metadata: license, adoption, age and documentation. Not a code audit — see the Safety scan above for what the skill file itself contains.

1 low