AgentBridge MCP
Provides read-only Git integration for observing repository state, changed files, and live Git status to coordinate handoffs between agents.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@AgentBridge MCPresume my last working state in this project"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
AgentBridge MCP
AgentBridge is a local MCP server and SQLite-backed shared working-state manager for handoffs between coding agents. Git remains authoritative for files and repository state; AgentBridge stores concise tasks, progress, decisions, commands, tests, checkpoints, and handoffs.
Architecture
Codex / Antigravity / other MCP client
| stdio MCP
v
AgentBridge (FastMCP)
tasks | progress | decisions | checkpoints
command/test metadata | read-only Git observer
|
v
SQLite WAL: %USERPROFILE%/.agentbridge/agentbridge.dbRelated MCP server: Geond Agent Protocol
Install and run
python -m venv .venv
.venv\Scripts\python -m pip install -e .[dev]
agentbridge status --path D:\Projects\example
.venv\Scripts\python -m agentbridge.serverSet AGENTBRIDGE_HOME to override the default global storage directory (useful in locked-down environments). The database is deliberately outside repositories.
MCP tools
register_project, get_project_state, start_task, get_current_task, update_progress, complete_task, record_decision, get_decisions, record_command, get_recent_commands, record_test_result, get_test_results, record_file_change, get_changed_files, get_git_state, checkpoint, handoff, and resume_task.
Workflow
Call resume_task when entering a project. For a new goal call start_task; record meaningful progress and decisions; record important tests; then call checkpoint and handoff before a planned switch. If a prior agent stops without either call, resume_task combines active task/progress with live Git state, commands and tests. Live Git state wins over a checkpoint and produces a stale-state warning.
CLI
agentbridge status, agentbridge resume, agentbridge checkpoint --agent codex, agentbridge tasks, and agentbridge history accept --path.
Security and limits
Likely credentials in persisted command summaries are redacted and text is size-limited. AgentBridge never reads .env and uses only read-only Git commands. It does not synchronize chat history or hidden reasoning; agents must record key progress to get the best handoff context.
Testing
.venv\Scripts\python -m pytest -qThe test suite creates temporary repositories and covers normal/reverse handoff, unexpected interruption without a checkpoint, stale-state detection, normalization, non-Git projects, and redaction.
This server cannot be deployed
Maintenance
Related MCP Connectors
Shared task board and knowledge base for AI coding agents Give your coding agents a shared task board and knowledge base, so the plan survives between sessions and across agents.
Durable agent-to-agent handoffs and shared scratchpad for multi-agent workflows.
Shared memory for coding agents. Stop re-explaining your codebase every session.
Git-backed platform for skills, tools, and context for AI agents
Related MCP Servers
- AlicenseNot gradedqualityAmaintenanceA coordination layer for coding agents that provides memorable identities, inbox/outbox messaging, searchable message history, and file lease management to prevent conflicts. Uses Git for human-auditable artifacts and SQLite for fast queries, enabling multiple agents to collaborate across projects without stepping on each other.2,156MIT
- AlicenseAqualityCmaintenanceLocal-first shared memory and coordination layer for AI coding agents, with repository evidence, reservations, handoffs, code graph context, and dashboard review backed by PostgreSQL/pgvector.303Apache 2.0
- FlicenseAqualityBmaintenanceEnables coding agents to track and coordinate project work through a shared SQLite ledger, including task plans, session ancestry, claims, work locations, blockers, and commits via MCP tools.9-
- AlicenseNot gradedqualityCmaintenanceGives AI coding agents persistent, branch-aware memory and a dependency-tracked task graph by storing decisions, lessons, and tasks as plain JSON and Markdown committed directly into the repository. Agents can record and fuzzy-search past decisions, dump instant project context, and create, claim, complete, and query tasks whose completion automatically unblocks downstream work.MIT