Skip to main content
Glama
deenesjakoruzh
chelslava

ContextTree MCP

by chelslava
OverviewSchemaSourceRelated ServersDiscussionsAdmin
Python
Local

:# ContextTree MCP 🌳

Tiefgehende semantische Codesuche für KI-Coding-Assistenten – unterstützt durch AST-Parsing und lokale Embeddings. 100 % offline.

Python License: MIT MCP README (RU)

ContextTree MCP ist ein lokaler Model Context Protocol-Server, der deinem KI-Coding-Assistenten strukturelles Verständnis einer Codebasis verleiht. Er verbindet zwei Welten:

  • tree-sitter zerlegt Quelldateien in einen AST und extrahiert logische Blöcke – Funktionen, Methoden und Klassensignaturen samt Docstrings.

  • sentence-transformers (all-MiniLM-L6-v2) bettet jeden Block – angereichert mit Dateipfad, Klassname, Methodenname und Docstring – in einen lokalen ChromaDB-Vektor-Store ein.

Das Ergebnis: Dein Assistent findet Code über die Bedeutung statt über Stichwörter, und jeder Suchtreffer enthält exakte Dateipfade und Zeilennummern.

🔒 Privatsphäre zuerst. Alles läuft auf deiner Maschine: Parsing, Embedding-Modell, Vektorindex. Keine Cloud-Aufrufe, keine Telemetrie, kein Code verlässt je den Rechner, auf dem er lebt.

Warum nicht einfache Textsuche?

grep und Volltextsuche gleichen Zeichenketten ab. Sie scheitern genau dort, wo Entwicklerinnen am häufigsten Hilfe brauchen:

Aufgabe

Textsuche

ContextTree MCP

„Wo validieren wir JWT-Tokens?“

❌ verlangt exaktes Raten von Stichwörtern

✅ trifft AuthService.verify_token(), auch wenn „validieren“ nirgends im Code vorkommt

Methode finden, deren Name umbenannt wurde

❌ bricht bei Umbenennung

✅ Die Bedeutung wird weiterhin Docstring und umgebenden Kontext enthalten

Definition von Verwendung unterscheiden

❌ unmöglich ohne Regex-Akrobatik

✅ AST-basierte Trennung von Deklarationen und Aufrufstellen

Präzise Standorte zurückgeben

⚠️ nur Zeile des Treffers

✅ Metadaten file, class, method, start_line, end_line

Kommentare / Strings / Import-Rauschen ignorieren

❌

✅ Es werden nur echte logische Einheiten indexiert

Related MCP server: Paparats MCP

Funktionen

  • 🌲 AST-basiertes Chunking – indiziert Funktionen, Methoden und Klassensignaturen via tree-sitter, nicht nach willkürlichen Textfenstern.

  • 🧩 Kontext-angereicherte logische Blöcke – jedes indizierte Dokument enthält Dateipfad, zugehörige Klasse, Methodenname, Docstring und Rumpf eingebettet; so landen Abfragen wie „Zahlungswiederholungslogik“ direkt bei der richtigen Methode.

  • ⚡ Inkrementelle Indizierung – SHA-256-Inhaltshashes pro Datei; nur geänderte/neue/gelöschte Dateien werden neu verarbeitet.

  • 🔍 Semantische Suche – natürliche Sprachabfrage → rankierte Code-Fragmente mit exakten Zeilenbereichen.

  • 📞 AST-Aufrufsuche – findet echte Aufrufstellen jedes Symbols (Funktion oder Klasse) und filtert Fehlalarme wie String-Literale oder Kommentare heraus.

  • 🗄️ Persistenter lokaler Index – ChromaDB liegt in .chroma/, übersteht Neustarts und wird nie in Git eingecheckt.

  • 🔌 Stdio-MCP-Transport – funktioniert mit Claude Desktop, OpenCode, Cursor, Cline oder jedem anderen MCP-kompatiblen Client.

Technologie-Stack

Ebene

Technologie

Sprache

Python 3.12+

Aufruf

TensorFlow Protokoll

in English

AST-Parsing

tree SDK (Stdio-Transport)

Vektordatenbank

chromadb (lokaler persistenter Modus)

Einbidwards

sentence-transformers · all-MiniLM-L6-v2

Installation

Erforderlich ist Python 3.12+. uv wird als schneller, moderner Paketmanager empfohlen:

git clone https://github.com/<your-org>/mcp-context-tree.git
cd mcp-context-tree

# Option A — uv (recommended): resolves dependencies from pyproject.toml, locks uv.lock
uv sync

# Option B — classic pip + venv
python -m venv venv && source venv/bin/activate   # Windows: venv\Scripts\activate
pip install -r requirements.txt

💡 Beim ersten Start wird das Einbettungsmodell (ca. 90 MB) heruntergeladen. PyTorch wird als Abhängigkeit von sentence-transformers installiert; die CPU-Variante reicht aus – keine GPU erforderlich.

Registrierung bei einem MCP-Client

Beispielkonfiguration (Claude Desktop / jeder Client mit Stdio-Unterstützung):

{
  "mcpServers": {
    "context-tree": {
      "command": "<path-to-venv>/bin/python",
      "args": ["-m", "context_tree"],
      "env": {}
    }
  }
}

(Unter Windows <path-to-venv>\Scripts\python.exe verwenden.)

Tools für den Assistenten

Tool

Signatur

Beschreibung

index_workspace

(directory_path: str)

Läuft durch das Projekt, erkennt per Hash geänderte Dateien und aktualisiert die ChromaDB-Sammlung inkrementell.

semantic_search

(query: str, limit: int = 5)

Durchsucht den indizierten Code semantisch. Liefert angereicherte Ausschnitte mit file, class, start_line, end_line.

find_ast_usages

(symbol_name: str)

AST-basiertes Auffinden echt durch echter Aufrufstellen (Instance) einer Funktion oder Klasse.

Typischer Arbeitsablauf:

1. index_workspace("D:/projects/my-app")
2. semantic_search("where do we handle payment retries", limit=8)
3. find_ast_usages("PaymentGateway.retry")

Unterstützte Sprachen

Sprache

Status

Python

✅ bei Veröffentlichung unterstützt

TypeScript / TSX

✅ bei Release unterstützt

JavaScript/JSX

✅ bei Release unterstützt

Go, Java, Rust …

🗺️ Roadmap – die Parser-Registry ist für Erweiterungen konzipiert

Dokumentation

  • 🏛️ ARCHITECTURE.md – Projektstruktur, ChromaDB-Datenschema, AST-Extraktionslogik, Entwurf der inkrementellen Indizierung.

  • 🇷🇺 README.ru.md – Dokumentation auf Russisch.

Roadmap

  • Watch-Modus – automatische Neuindizierung bei Dateiänderungen (FS-Watcher)

  • Weitere Sprachen via tree-sitter-Bindings

  • Hybride Suche (BM25 + Vektor-Reranking)

  • Aufrufgraph-basiertes Ranglistung („Wer ruft das?“)

  • PyPI-Veröffentlichung (context-tree-mcp)

Mitwirken

Alle und PRs sind willkommen. Bitte halte Änderungen mit ARCHITECTURE.md konsistent.

Lizenz

MIT – siehe LICENSE.

Available Tools

4 tools
find_ast_usagesA

AST-based lookup of real call sites / instantiations of a function or class. Filters out string literals, comments, and non-call occurrences.

ParametersJSON Schema
NameRequiredDescriptionDefault
limitNo
symbol_nameYes
directory_pathNo.

Output Schema

ParametersJSON Schema
NameRequiredDescription
resultYes

TDQS

A3.6/5.0
Behavior3/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

With no annotations supplied, the description carries the full behavioral burden. It does disclose a valuable filtering behavior: "Filters out string literals, comments, and non-call occurrences." However, it does not mention whether indexing is required, how missing symbols are handled, or any result behavior beyond being AST-based.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness5/5

Is the description appropriately sized, front-loaded, and free of redundancy?

The description is a single focused sentence that communicates the mechanism, target resource, and exclusion behavior without redundancy. Every clause adds useful information.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness3/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

The tool is simple enough that the core lookup behavior is adequately described, and an output schema exists. The main gaps are the lack of usage guidance around semantic_search/index_workspace and the absence of practical details about scoping and limit behavior, which leaves the description merely adequate rather than complete.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters2/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

The schema has 0% description coverage, so the description should compensate. It adds meaning for symbol_name by stating it refers to "a function or class," but it gives no guidance on directory_path or limit, such as how the directory is searched or what the limit controls. This leaves non-obvious semantics undocumented.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description uses a specific verb plus precise resource: "AST-based lookup of real call sites / instantiations of a function or class." It also clarifies what is excluded, which distinguishes it from string or semantic search. This is far more informative than the bare tool name.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines3/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

The AST-based wording implicitly indicates this is for exact code-structure matching rather than fuzzy search, and the sibling tools suggest a semantic-search alternative. However, the description does not explicitly state when to use this tool versus semantic_search or index_workspace, nor does it mention any exclusions or preconditions.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

go_to_definitionA

Finds the exact definition/declaration location of a symbol (function, method, class, struct, trait, interface) across workspace files.

ParametersJSON Schema
NameRequiredDescriptionDefault
limitNo
symbol_nameYes
directory_pathNo.

Output Schema

ParametersJSON Schema
NameRequiredDescription
resultYes

TDQS

A3.5/5.0
Behavior3/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

With no annotations provided, the description carries the full burden of behavioral disclosure. It adds meaningful context by stating the search is workspace-wide and targets exact definition/declaration locations. Still, it leaves important behavioral details undisclosed, such as whether an index must exist first, how missing or ambiguous symbols are handled, and whether the operation is purely read-only.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness5/5

Is the description appropriately sized, front-loaded, and free of redundancy?

The description is a single, front-loaded sentence with no filler or redundant restatement of the tool name. It conveys the core action, target, and scope efficiently while listing relevant symbol categories without over-elaborating.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness2/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

Although an output schema exists and the basic action is clear, the description lacks usage guidance relative to its three sibling tools and fails to explain two of three parameters. With no annotations, this is too thin to fully support an agent in choosing and invoking the tool correctly across realistic scenarios.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters2/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema description coverage is 0%, so the description must compensate for bare parameter schemas. It partially clarifies symbol_name by listing supported symbol types, but it says nothing about the format of symbol_name (e.g., fully qualified vs simple name, case sensitivity). The limit and directory_path parameters are completely unaddressed in the description, leaving their semantics to inference from the schema titles and defaults.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description uses a specific verb ('Finds') with a precise resource: the exact definition/declaration location of symbols, listing supported symbol categories and clarifying the search scope ('across workspace files'). This clearly distinguishes it from sibling tools like find_ast_usages or semantic_search, which target usages or semantically similar matches rather than exact definitions.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines3/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

The phrasing 'exact definition/declaration location' implies the tool is appropriate when precise definitions are needed, rather than usages or semantic matches. However, it does not explicitly state when to prefer this tool over semantic_search or find_ast_usages, nor does it mention any exclusions or prerequisites such as requiring an indexed workspace.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

index_workspaceA

Walks the project, detects changed files by SHA-256 hash, and incrementally updates the persistent local ChromaDB vector store.

ParametersJSON Schema
NameRequiredDescriptionDefault
directory_pathNo.

Output Schema

ParametersJSON Schema
NameRequiredDescription
resultYes

TDQS

A3.9/5.0
Behavior4/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

No annotations are provided, so the description carries the full burden of behavioral disclosure. It covers that the operation walks the project, uses SHA-256 hashing for change detection, and mutates a persistent local ChromaDB store. This is meaningful behavioral context beyond 'index workspace' though it does not mention potential costs, permissions, or failure modes.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness5/5

Is the description appropriately sized, front-loaded, and free of redundancy?

A single sentence with no filler. It front-loads the core action and packs in the hashing, incrementality, and storage target without becoming verbose.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness4/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

For a tool with one optional parameter and an output schema, the description covers the essential behavior: what is walked, how changes are detected, and what is updated. It is reasonably complete, though it could add an explicit note about running after source changes or before semantic_search.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters2/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

There is only one parameter and schema description coverage is 0%, so the description needs to clarify the parameter's meaning. It never mentions directory_path or the default of '.', forcing the agent to rely on the schema's 'Directory Path' title and default value. The description does not compensate for the low schema coverage.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description names a specific action sequence ('Walks the project', 'detects changed files', 'updates the vector store') and a concrete resource ('persistent local ChromaDB vector store'). It is clearly distinct from the sibling tools semantic_search and find_ast_usages, which do search and AST lookup rather than indexing.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines3/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

The phrase 'incrementally updates' and 'detects changed files' implies the tool is meant to keep the index fresh after edits, but the description never explicitly says when to run it relative to semantic_search or find_ast_usages. It also does not mention exclusions or alternatives.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

  1. 2 tool updatesv0.4.0
    • Addedgo_to_definition
    • Changedsemantic_search2 fields changed
      • addedInput schema / properties / mode
        Added value: +{
+  "default": "hybrid",
+  "enum": [
+    "hybrid",
+    "semantic",
+    "keyword"
+  ],
+  "title": "Mode",
+  "type": "string"
+}
      • addedInput schema / properties / rerank
        Added value: +{
+  "default": false,
+  "title": "Rerank",
+  "type": "boolean"
+}
  2. 3 tool updatesv0.1.0
    • First observedfind_ast_usages
    • First observedindex_workspace
    • First observedsemantic_search
Deploy Server
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

TDQS

A3.9/5.0

Scored across 4 tools

Disambiguation5/5

Each tool targets a distinct part of code navigation: semantic search vs. exact definitions vs. AST usages vs. index maintenance. There is no realistic confusion between their purposes.

Naming Consistency4/5

Three of four tools use imperative verb-first names (find_ast_usages, index_workspace, go_to_definition), and all are snake_case. semantic_search breaks the pattern slightly by starting with an adjective instead of a verb, but the naming remains readable and predictable.

Tool Count5/5

With 4 tools, the server is well-scoped and each tool earns its place in the code indexing/search/navigation workflow. There is no redundancy or bloat.

Completeness4/5

The core lifecycle is covered: index the workspace, search semantically, go to definitions, and find AST usages. Minor gaps such as no explicit index status or reset operation prevent a perfect score, but agents can work around them.

Maintenance

ActivitySlowing
ResponsivenessResponsive

Resources

Related MCP Connectors

Related MCP Servers

  • CodeGrok MCP

    dondetir
    A
    license
    Not graded
    quality
    D
    maintenance
    Enables semantic code search for AI assistants by indexing codebases with embeddings and Tree-sitter, returning relevant snippets via natural language queries.
    15
    MIT

  • Paparats MCP

    IBazylchuk
    A
    license
    Not graded
    quality
    B
    maintenance
    Provides AI coding assistants with deep, semantic understanding of local codebases via AST-aware chunking, cross-repo symbol graphs, and architectural memory, enabling context-aware code search and dependency tracing.
    11
    MIT

  • smart-coding-mcp

    amalikn
    A
    license
    A
    quality
    D
    maintenance
    Enables AI assistants to perform intelligent semantic code search across codebases using local AI embeddings for meaning-based retrieval.
    6
    35 npm
    MIT