A/MCL — Agent/Multi-Coding-agent Context Layer. MCP server for automatic context persistence across AI coding agents.
Project description
A/MCL — Agent/Multi-Coding-agent Context Layer
Zero-intervention context persistence across AI coding agents.
When you hit a rate limit on one agent (Antigravity, Cursor, Claude Code) and switch to another, A/MCL ensures the new agent automatically has access to the complete conversation history, file changes, reasoning chains, and project state. No commands. No manual handoff. It just works.
How It Works
A/MCL is an MCP (Model Context Protocol) server. MCP-compatible agents discover it automatically at launch and gain access to shared project context via standard tools, resources, and prompts.
┌──────────────┐ stdio ┌────────────────┐ SQLite ┌──────────┐
│ AI Agent │ ◄────────────► │ A/MCL Server │ ◄────────────► │ Context │
│ (Antigravity,│ │ (FastMCP) │ │ Store │
│ Cursor, etc)│ │ │ │ (~/.amcl)│
└──────────────┘ └────────────────┘ └──────────┘
- Install once →
pip install amcl-server(orpip3 install) - Setup →
amcl-server setup(creates DB, shows MCP config) - Use your agents normally → Context is automatically shared
- Switch agents → New agent picks up exactly where the last one left off
Quick Start
# Install universally
pip install amcl-server # or pip3 install amcl-server
# Initialize and Auto-Register
amcl-server setup
# Verify integrations
amcl-server check
# Check status
amcl-server status
Upgrading an Existing Installation
pip install --upgrade amcl-server
amcl-server setup
amcl-server check
The setup command automatically detects all installed AI agents and IDE extensions on your system and registers A/MCL directly into their settings. Once you run setup, you are completely done.
Supported Agents (Auto-Detected)
A/MCL automatically integrates with:
- Cursor (
~/.cursor/mcp.json) - Claude Desktop / Antigravity (
Library/Application Support/Claude/claude_desktop_config.json) - Amp (
~/.amp/mcp.json) - Roo / Cline (VSCode & Cursor instances)
- Generic MCP clients (
~/.mcp/config.json)
If your agent isn't automatically found, the setup command will print the exact JSON snippet you can manually paste into its MCP configuration.
What Gets Shared
| Category | Data |
|---|---|
| Conversation | All messages between user and agents, with agent attribution |
| File Changes | Files created, modified, deleted — tracked automatically |
| Tasks | Current goal, active tasks, completed work, blockers |
| Decisions | Design decisions with rationale and alternatives considered |
| Agent History | Which agents worked on what, session timestamps |
MCP Tools
Agents can invoke these tools:
| Tool | Description |
|---|---|
context_get_current |
Full project context snapshot |
context_update |
Append new conversation/file/decision data |
context_query |
Search context history |
context_get_files_changed |
Files changed since last agent switch |
context_get_conversation |
Recent N messages |
context_get_reasoning |
Decision log |
context_add_decision |
Log a design decision |
context_mark_complete |
Mark a task as completed |
context_add_blocker |
Record a blocking issue |
MCP Resources
Read-only data endpoints:
context://current-project— Complete snapshotcontext://conversation— Message historycontext://files— File state & changescontext://reasoning— Decision logcontext://tasks— Task statecontext://agents— Agent session history
MCP Prompts
Pre-built prompts for context injection:
- Project Context — Complete project overview
- Recent Work — Summary of last session
- Continuation — "Pick up where we left off"
- Decisions Log — Key decisions and rationale
Environment Variables
| Variable | Default | Description |
|---|---|---|
AMCL_DATA_DIR |
~/.amcl |
Where context database lives |
AMCL_LOG_LEVEL |
WARNING |
Logging verbosity |
AMCL_AGENT_NAME |
unknown |
Name of the connecting agent |
AMCL_PROJECT_DIR |
cwd |
Project directory override |
Privacy
All data stays local on your machine in ~/.amcl/amcl.db. Nothing is ever sent to external servers. No telemetry, no analytics.
License
MIT
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Filter files by name, interpreter, ABI, and platform.
If you're not sure about the file name format, learn more about wheel file names.
Copy a direct link to the current filters
File details
Details for the file amcl_server-1.0.4.tar.gz.
File metadata
- Download URL: amcl_server-1.0.4.tar.gz
- Upload date:
- Size: 21.7 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.2.0 CPython/3.11.7
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
17aa6666cab34652fa5c9346187457e558898b7b4dfc1880aabc466590cfd1e1
|
|
| MD5 |
503b97ba952209fef1e0fb3e4599d70b
|
|
| BLAKE2b-256 |
c5bc4c227b3abb6d1f94b5aaf9c2138252fe6262c9fc8ddc2934f41398db4064
|
File details
Details for the file amcl_server-1.0.4-py3-none-any.whl.
File metadata
- Download URL: amcl_server-1.0.4-py3-none-any.whl
- Upload date:
- Size: 24.5 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.2.0 CPython/3.11.7
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
63f02ee4f6ed18e706b1090ec5dab048168377ba06051819e52ab44c90e8a112
|
|
| MD5 |
775524d9782ed08d847ae8fbadb66340
|
|
| BLAKE2b-256 |
dc6c6d542fae89bf40ea055080be790c6713af4c88834b2a4abc6905491863fc
|