Skip to main content
Glama
Sitten-Tokyo

Sippion

by Sitten-Tokyo
OverviewSchemaSourceRelated ServersDiscussionsAdmin
Rust
Local

Sippion

Release CI MCP Registry License Rust

English | 日本語

Use less AI context. Make API tokens — and subscription usage allowances — last longer.

Sippion is a local, read-only MCP server that helps AI coding agents consume fewer tokens. Instead of letting an agent open large parts of a repository, Sippion's single tool, repo_context, returns a small set of relevant, bounded source excerpts first.

This reduces unnecessary context per task. It can lower usage for API-based workflows, but that is not the only benefit: Sippion can also help subscription-based AI coding plans last longer by reducing how much context the agent needs to consume.

How it works

AI coding agent
    ↓ ask for focused repository context
Sippion repo_context
    ↓ return a few relevant excerpts
AI reads only the files it actually needs

Example:

repo_context {"q":"authentication token validation"}

Sippion uses bounded lexical, structural, and semantic retrieval while staying local and read-only.

Related MCP server: git-context-mcp

Install

macOS / Linux

curl -fsSL --proto '=https' --proto-redir '=https' --tlsv1.2 https://raw.githubusercontent.com/Sitten-Tokyo/Sippion/75d6b27e83b86bec00297cd5b5c05bb014e16904/scripts/bootstrap.sh | sh

Windows PowerShell

irm https://raw.githubusercontent.com/Sitten-Tokyo/Sippion/75d6b27e83b86bec00297cd5b5c05bb014e16904/scripts/bootstrap.ps1 | iex

Then restart your AI client. sippion setup registers Sippion for Codex, Claude Code, Antigravity, and OpenCode.

Useful commands:

sippion setup
sippion doctor
sippion uninstall
sippion mcp --root-auto

Why Sippion

  • Fewer tokens: gives the agent focused repository context instead of broad file reads.

  • Useful beyond API billing: can help subscription usage allowances last longer too.

  • Local and private: no network access while serving repository context.

  • Read-only: never modifies the repository or executes repository code.

  • No persistent index: retrieval state stays in RAM.

  • Small surface area: one MCP tool, repo_context.

Sippion is a repository-context tool, not a compiler or language server. Structural parsing currently covers Rust, Python, JavaScript/TypeScript, Go, Java, C#, C, and C++.

Safety

Sippion does not run repository code, build scripts, compilers, LSP servers, or shell commands during retrieval. Repository text is treated as untrusted data, and high-confidence secrets are redacted before model output.

See Security and trust boundary for details.

Documentation

Credits

The compact efficiency behavior is inspired by Ponytail and i-have-adhd. License notes are in Third-party notices.

This server cannot be deployed

One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

Maintenance

ActivityMaintained
ResponsivenessResponsive

Resources

Related MCP Connectors

Related MCP Servers

  • Local Code MCP Server

    udittripathi
    F
    license
    B
    quality
    D
    maintenance
    Provides LLMs with safe, read-only access to local codebases for searching, reading files, and finding function definitions. All source code remains local, ensuring privacy while enabling AI assistants to explore project structures and functionality.
    4
    -

  • Context MCP

    geq1fan
    A
    license
    Not graded
    quality
    D
    maintenance
    Provides AI agents with secure, read-only file system access to analyze and understand project codebases, enabling multi-repository context aggregation and cross-project code tracing.
    5
    MIT

  • token-context-mcp

    MCK564
    A
    license
    A
    quality
    B
    maintenance
    Enables coding agents to retrieve token-budgeted, source-hashed code context from explicitly registered local repositories via read-only MCP tools, reducing broad repository crawling while providing structure, symbols, and impact slices.
    9
    MIT