Skip to content
Sephera Docs

Sephera

Sephera

Section titled “Sephera”

Sephera is a Rust workspace for codebase inspection. It currently focuses on four practical workflows:

  • loc for fast, language-aware line counting
  • context for deterministic Markdown or JSON context packs with AST compression
  • graph for multi-language dependency graph analysis
  • mcp for built-in MCP server agent integration

The current docs reflect the v0.5.x release line.

The project is intentionally narrow in scope. It does not try to be an AI agent framework or a hosted service. The goal is to provide reliable local analysis primitives that fit naturally into review, debugging, and prompting workflows.

Why it exists

Section titled “Why it exists”

Modern code workflows need two kinds of signals:

  • trustworthy repository metrics
  • focused context bundles that are small enough to fit inside real prompt budgets

Sephera provides both without requiring a server, a browser extension, or a provider-specific integration.

Current capabilities

Section titled “Current capabilities”
  • Fast loc analysis with per-language totals, table output, and elapsed-time reporting
  • Deterministic context packs with focus-path prioritization, Git diff awareness, and approximate token budgeting
  • Tree-sitter AST compression reducing token usage by 50-70% for 8 supported languages
  • Dependency graph generation with cycle detection and exports to Markdown, JSON, XML, and DOT
  • Built-in MCP server for direct integration with AI agents like Claude Desktop
  • URL mode for direct analysis of cloneable repo URLs and GitHub/GitLab tree URLs
  • Repo-level defaults and named profiles through .sephera.toml
  • Export to Markdown for human copy-paste workflows and JSON for automation
  • Generated language metadata sourced from config/languages.yml
  • Byte-oriented scanning with newline portability across LF, CRLF, and classic CR
  • Benchmark and fuzzing infrastructure to keep behavior stable over time

Quick examples

Section titled “Quick examples”

These examples assume sephera is installed and available on your PATH.

Count lines of code in the current repository:

Terminal window
sephera loc --path .

Analyze a remote repository directly:

Terminal window
sephera loc --url https://github.com/Reim-developer/Sephera

Build a focused context pack and export it to JSON:

Terminal window
sephera context --path . --focus crates/sephera_core --format json --output reports/context.json

Compress context excerpts to reduce LLM token usage:

Terminal window
sephera context --path . --compress signatures

Start the MCP server to let AI agents call Sephera directly:

Terminal window
sephera mcp

Build a review pack from recent Git changes:

Terminal window
sephera context --path . --diff HEAD~1 --budget 32k

Run the same review flow on a remote checkout:

Terminal window
sephera context --url https://github.com/Reim-developer/Sephera --ref master --diff HEAD~1 --budget 32k

List configured profiles for the current repository:

Terminal window
sephera context --path . --list-profiles

Terminal demos

Section titled “Terminal demos”
sephera loc language-aware repository totals
Terminal demo of sephera loc rendering a table report.
sephera context deterministic context bundles for people, tools, and Git review flows
Terminal demo of sephera context building a structured context pack.

The demos above are illustrative captures of the CLI workflows described throughout the docs.

Where to go next

Section titled “Where to go next”