
Cut code-reading token costs by up to 99%
Most AI agents explore repositories the expensive way: open entire files > skim thousands of irrelevant lines > repeat.
Potencia tu agente
Instalación del publicador. No es un endpoint de PlugMCP.
Host a copy with our keys and logs.
Aún no hay una versión empaquetada en el registro. Sigue el README del repositorio.
¿URL remota o un comando en tu máquina? En qué se diferencian las dos instalaciones
Ejecutar en PlugMCP
A new server you create. It is not this listing's official endpoint. Use the publisher install when you want their URL and their credentials.
AbrirDel publicador

Most AI agents explore repositories the expensive way: open entire files > skim thousands of irrelevant lines > repeat.
Structured code retrieval for AI agents — index once with tree-sitter, query symbols precisely via MCP. Cut code-reading token costs by up to 99%.
ASTral indexes a codebase once and lets agents retrieve only the exact symbols they need — functions, classes, methods, constants — with byte-level precision.
Index once. Query cheaply forever. Precision context beats brute-force context.
Stop dumping files into context windows. Start retrieving exactly what the agent needs.
ASTral indexes a codebase once using tree-sitter AST parsing, then allows MCP-compatible agents (Claude Desktop, VS Code, Google Antigravity, and others) to discover and retrieve code by symbol instead of brute-reading files.
Every symbol stores:
Full source is retrieved on demand using O(1) byte-offset seeking.
Repo: geekcomputers/Python
Size: 338 files, 1,422 symbols indexed
Task: Locate calculator / math implementations
Cost scales with tokens. Latency scales with irrelevant context.
ASTral turns search into navigation.
Agents waste money when they:
ASTral provides precision context access:
Agents do not need larger context windows. They need structured retrieval.
~/.code-index/){file_path}::{qualified_name}#{kind}
Examples:
src/main.py::UserService.login#methodsrc/utils.py::authenticate#functionIDs remain stable across re-indexing when path, qualified name, and kind are unchanged.
index_folder: { "path": "/path/to/project" }
index_repo: { "url": "owner/repo" }
get_repo_outline: { "repo": "owner/repo" }
get_file_outline: { "repo": "owner/repo", "file_path": "src/main.py" }
search_symbols: { "repo": "owner/repo", "query": "authenticate" }
get_symbol: { "repo": "owner/repo", "symbol_id": "src/main.py::MyClass.login#method" }
search_text: { "repo": "owner/repo", "query": "TODO" }
Every tool response includes a _meta envelope with timing, token savings, and cost avoided:
"_meta": {
"timing_ms": 4.3,
"tokens_saved": 48153,
"total_tokens_saved": 1280837,
"cost_avoided": { "claude_opus": 0.7223, "gpt5_latest": 0.4815 },
"total_cost_avoided": { "claude_opus": 19.21, "gpt5_latest": 12.81 }
}
total_tokens_saved and total_cost_avoided accumulate across all tool calls and persist to ~/.code-index/_savings.json.
* .h is parsed as C++ first, then falls back to C when no C++ symbols are extracted.
See Language Support for full semantics.
Built-in protections:
.env, *.pem, etc.)See Security for details.
dotnet tool install -g ASTral
Then run:
astral
git clone https://github.com/Atypical-Consulting/ASTral.git
cd ASTral
dotnet build
dotnet run --project src/ASTral
ASTral can be configured via environment variables, a .astralrc JSON config file, or both.
.astralrc)Place a .astralrc JSON file in your project directory or home directory (~/.astralrc). Project-level values override home-level values. Environment variables override both.
{
"storage_path": "/custom/index/path",
"log_level": "DEBUG",
"max_index_files": 5000,
"extra_extensions": ".vue:javascript,.svelte:javascript",
"excluded_patterns": ["*.generated.cs", "*.Designer.cs"]
}
Both index_repo and index_folder accept a force parameter to bypass incremental cache and perform a full re-index:
{ "path": "/path/to/project", "force": true }
Index tool responses include timing breakdown:
{
"parse_time_ms": 120,
"save_time_ms": 45,
"total_time_ms": 165
}
Set ASTRAL_WATCH=true to enable automatic background re-indexing when files change in indexed local folders. The watcher uses debounced detection (500ms) and a semaphore to prevent concurrent indexing.
ASTral is a .NET rewrite of jCodeMunch-MCP by J. Gravelle, originally written in Python.
This derivative work includes the following modifications:
This repository is free for non-commercial use under the original dual-use license by J. Gravelle. Commercial use requires a paid commercial license — see LICENSE for full terms.
For commercial licensing inquiries, contact: [email protected] | https://j.gravelle.us
index_folder/index_repo MCP toolsASTRAL_WATCH file-watching to remote/CI-triggered re-indexing.NET global tool and install.shTrack progress and propose ideas via the open issues.
Part of a suite of Model Context Protocol servers by Atypical Consulting:
Contributions are welcome. Open an issue first to discuss any significant change.
git checkout -b feat/my-feature)git commit -m 'feat: ...')Built by Atypical Consulting. We also make NuGetKeep, a self-hosted NuGet server with supply-chain quarantine.
Structured code retrieval for AI agents — index once with tree-sitter, query symbols precisely via MCP. Cut code-reading token costs by up to 99%.
Herramientas
Aún no hay lista de herramientas. Los servidores remotos la reciben tras la primera comprobación de estado.
Preguntas frecuentes
Copia en esta página la configuración stdio para Cursor, Claude Desktop, Claude Code o VS Code. Usa la instalación del publicador, no un proxy de PlugMCP.
Este registro no indica ninguna variable de entorno obligatoria.
ASTral se ejecuta localmente por stdio.
Última comprobación: 2026-11-14. Structured code retrieval for AI agents — index once with tree-sitter, query symbols precisely via MCP. Cut code-reading token costs by up to 99%.