AgentiBridge — The Workforce Network

Three roles in one MCP server.

Bridge plays three parts simultaneously. An artwork that shows only one is wrong — and so is a marketing pitch.

🔎

Workforce memory

Every Claude Code transcript auto-indexed. Embedded with pgvector. Semantically searchable from any MCP client. Ask "what was I working on Tuesday at 8?" and get the session back, ready to resume with one shell command.

For the operator

🤝

A2A registry

Specialised agents register themselves, heartbeat with status, and discover peers by capability across machines. spec-01 needs a database refactor done? It calls find_agents, finds spec-03, hands off the work. The fleet stops waiting on you.

For the agents

🔄

Project handoff

Seed a primed conversation in any other project's agent with structured context — summary, decisions, next steps. One agent picks up exactly where another left off, even on a different machine.

For collaboration

agentibridge search

Ask your AI's history anything.

From any MCP client — Claude Code, claude.ai, VS Code, ChatGPT, Grok, your own automation — semantic search works the same way. You don't have to remember exact words. You don't have to remember which project. You ask in natural language, Bridge returns ranked sessions with the resume command.

How it works: the collector watches ~/.claude/projects/, incrementally embeds new transcripts as conversation-turn chunks via the OpenAI-compatible LLM API, stores vectors in Postgres+pgvector with an HNSW index, and serves cosine-similarity queries deduplicated per session.

From the shell: agentibridge search "<query>" spawns a headless Opus one-shot that reasons over your sessions, history, memory, and git log, streams every tool call live to your terminal, and prints a summary with a claude --resume <session_id> footer.

$ agentibridge search recon

"what was I working on Tuesday around 20:00?"

match 0.91 8425afcc email-template — debugging Jinja inheritance

match 0.84 7242ddf3 mcp-config — wiring agentibrain server

match 0.79 9705e5e1 build-fix — pre-commit hook breakage

→ resume: claude --resume 8425afcc

search_semantic search_sessions agent_search restore_session generate_summary

30 MCP tools across 6 surfaces.

Every tool is an MCP function call from your client. Browse the surface that matches your job.

Foundation

list_sessionsBrowse all indexed sessions

get_sessionRetrieve full session transcript

get_session_segmentFetch specific conversation segment

get_session_actionsList tool calls + actions taken

search_sessionsKeyword search across sessions

collect_nowTrigger immediate transcript indexing

AI-Powered

search_semanticNatural language search with embeddings

generate_summaryAI-generated session summaries

agent_searchHeadless one-shot recon over sessions + git + memory

Dispatch & Plans

restore_sessionResume a previous session

dispatch_taskFire-and-forget host-local Claude one-shot

get_dispatch_jobCheck job status + results

list_dispatch_jobsAll jobs with status filter

plan_taskCreate implementation plan

execute_planRun a plan as a coding job

get_dispatch_planRetrieve plan content

list_dispatch_plansAll plans with status filter

Knowledge Catalog

list_memory_filesBrowse memory files by project

get_memory_fileRead specific memory file contents

list_plansView generated implementation plans

get_planRead full plan details

search_historySearch across prompt history

Cross-Project Handoff

list_handoff_projectsDiscover target projects for handoff

handoffSeed a primed conversation in any project

A2A Agent Registry

register_agentAnnounce an agent to the fleet

heartbeat_agentKeep-alive ping with status

deregister_agentRemove an agent on shutdown

list_agentsDiscover all registered agents

get_agentFetch a specific agent's metadata

find_agentsFilter by capability across the fleet

Three workflows Bridge unlocks.

📚

"Resume my Tuesday work."

Operator · semantic recall

You opened a session two days ago, fixed a tricky bug in the email-template renderer, then closed the terminal. Today you want to pick up where you left off. From any MCP client: "find the session about email templates from Tuesday". Bridge returns the ranked match with a claude --resume command. You're back in flow in one prompt.

🔗

"Hand this off to spec-03."

Agent-to-agent collaboration

spec-01 is mid-conversation with you about a feature. The next step needs deep database refactoring — spec-03's specialty. spec-01 calls find_agents capability=db-refactor, gets spec-03's endpoint, calls handoff with a structured summary. spec-03 picks up the work in its own project, with full context. You watch from the operator chair.

🌐

"Drive my fleet from VS Code."

MCP everywhere

Bridge speaks MCP. VS Code Copilot, Cursor, claude.ai, ChatGPT, Grok, your own shell automation — they all see Bridge as a server full of tools. The same search_semantic call works from a terminal, from a chat, from a CI step. One control plane, every client you already use.

AgentiBridge.