Skip to main content

Combined Context — One MCP server. Every AI tool. No re-explaining.

Project description

one-context — Combined Context

One MCP server. Every AI tool. No re-explaining.

Architecture Diagram

Claude Code, Cline, Antigravity, Codex, Ollama — they all point to the same place.
You explain your project once. Every tool knows it forever.

No API keys needed. No cloud. No accounts. 100% local.

The Architecture

Every AI tool keeps its own internal memory. That doesn't change. But they all read and write to one shared root:

graph TD
    %% Define styles
    classDef rootNode fill:#0d1117,stroke:#58a6ff,stroke-width:3px,color:#c9d1d9,font-size:16px,font-weight:bold;
    classDef aiNode fill:#161b22,stroke:#3fb950,stroke-width:2px,color:#c9d1d9,font-size:14px;
    classDef bucketNode fill:#21262d,stroke:#8b949e,stroke-dasharray: 5 5,color:#c9d1d9;

    %% Nodes
    A([one-context MCP]):::rootNode
    
    C1[Claude Code]:::aiNode
    C2[Cline]:::aiNode
    C3[Antigravity]:::aiNode
    C4[Codex]:::aiNode
    
    B1[(WHAT\nProject Scope)]:::bucketNode
    B2[(DONE\nHistory)]:::bucketNode
    B3[(NOW\nCurrent Task)]:::bucketNode
    B4[(MAP\nKey Files)]:::bucketNode

    %% Connections
    C1 <-->|reads/writes| A
    C2 <-->|reads/writes| A
    C3 <-->|reads/writes| A
    C4 <-->|reads/writes| A
    
    A --- B1
    A --- B2
    A --- B3
    A --- B4

Install

With uv installed, you don't even need to download this repository. Your AI tools will fetch it automatically from PyPI!

Connect Your AI Tools

Option A: Command/stdio (recommended)

Works with Claude Desktop, Cline, Codex, and any MCP client. Just add this to your MCP settings file:

{
  "mcpServers": {
    "one-context": {
      "command": "uvx",
      "args": ["one-context", "stdio"]
    }
  }
}

No server to start. The AI tool launches it automatically.

Option B: HTTP/SSE (for network setups)

uvx one-context serve

Then point any MCP client to http://localhost:7337/sse.

The Four Buckets

Bucket Contains Updated when...
WHAT Project description, stack, architecture, constraints Project-level info changes
DONE Decisions made, files changed, problems solved Any tool finishes a task
NOW Current task, current state, what's in progress A new task starts
MAP Important file paths and what they do AI discovers key files

Usage in any tool

Start a session:

"Load context from one-context for my-project"

Finish a session:

"Update one-context with what we just did"

Register important files:

"Use ctx_map to register src/main.py as the entry point"

Search across all your projects:

"Search one-context for 'SQLite lock error'"

MCP Tools (6 total)

Tool Description
ctx_get(project) Get the full WHAT/DONE/NOW/MAP snapshot + git info
ctx_update(project, session_summary, tool_name) Merge a session update into context
ctx_map(project, files) Register important files manually
ctx_search(query) Search across ALL projects' context and history
ctx_reset(project) Wipe project context to empty
ctx_list() List all tracked projects

Projects are auto-created on first ctx_update — you don't need ctx init.

Advanced Features

MAP — Important Files

When an AI tool discovers that src/main.py is the entry point, it saves those paths to the MAP bucket. The next AI tool instantly knows which files matter without scanning the entire codebase.

Git Branch Awareness

Link a project to its git repo:

ctx init my-project --path /path/to/repo

Now ctx_get automatically includes the current branch name, last 5 commits, and uncommitted changes.

Cross-Project Search

ctx search "SQLite"

Searches across ALL projects' context (what/done/now/map) and update history. Find how you solved a problem before.

Summarization Modes

Mode Setup API Key?
Local (default) Nothing No
Ollama CTX_OLLAMA_MODEL=llama3.2 No
Claude ANTHROPIC_API_KEY=... Yes
OpenAI / Groq / Together OPENAI_API_KEY=... Yes

Priority: Ollama > Claude > OpenAI > Local fallback. If any provider fails, it silently falls back. Local always works.

CLI Commands

Command Description
ctx stdio Run via stdio (for MCP clients)
ctx serve Start HTTP/SSE server
ctx init <project> [--path /repo] Initialize with optional git repo
ctx status [project] Show context + git info
ctx search <query> Search across all projects
ctx reset <project> Reset project context
ctx list List all projects
ctx delete <project> Delete a project

License

MIT

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

one_ctx-0.1.0.tar.gz (16.2 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

one_ctx-0.1.0-py3-none-any.whl (18.5 kB view details)

Uploaded Python 3

File details

Details for the file one_ctx-0.1.0.tar.gz.

File metadata

  • Download URL: one_ctx-0.1.0.tar.gz
  • Upload date:
  • Size: 16.2 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.10.0

File hashes

Hashes for one_ctx-0.1.0.tar.gz
Algorithm Hash digest
SHA256 8c5dd1795ff0be4c939e61ccf9582bc5acd8f0aa425585fd9d1ea28aa8954cc1
MD5 db9d365fdcc04227bce91683e0eca970
BLAKE2b-256 c147db7a71f384f371dd22dd8c5035f91c870674aebfa58bc8315cdc544a0030

See more details on using hashes here.

File details

Details for the file one_ctx-0.1.0-py3-none-any.whl.

File metadata

  • Download URL: one_ctx-0.1.0-py3-none-any.whl
  • Upload date:
  • Size: 18.5 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.10.0

File hashes

Hashes for one_ctx-0.1.0-py3-none-any.whl
Algorithm Hash digest
SHA256 4950bd00474c387ddb5b1bbfbd89e34f39f4d6073da4adbb151e67b2ac5096b0
MD5 05cbaabe6ed69f61c5b5a6128e22cc6b
BLAKE2b-256 bffe8d9f1463c73c5899f8ddfb49c7f31d9d3b3128da17847cd7f2d1ebeebe3f

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page