BirdEye
Click on "Install 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., "@BirdEyescan all harnesses and show me the inventory"
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.
🦅 BirdEye
One hub, every bird. Local-first mission control for all your AI agent harnesses — Claude Code, Claude Desktop, Codex, opencode, OpenClaw, Zcode, Gemini CLI, Cursor, Continue.
The problem
If you use more than one AI agent harness, you know this pain:
Every harness is an island. Each has its own sessions, memories, MCP servers, plugins, skills, and routines. Nothing is shared.
Memory doesn't travel. What Claude Code learned about you yesterday, Codex re-asks today. Import tools — where they exist — are per-harness and incompatible.
Credentials multiply. Giving agents access to GitHub means configuring auth in every single harness. N harnesses = N copies of every token.
Zero visibility. Which harness can touch what? How many tokens has each burned? Nobody can tell you.
No shared work. You can't assign a task from one place and have different harnesses pick it up.
Related MCP server: Bastion
What BirdEye does
BirdEye is a daemon on 127.0.0.1 + an interactive dashboard + one MCP gateway that every harness joins. Free, MIT, no cloud, no telemetry — your data never leaves your machine.
Pillar | What you get |
🗺️ Unified inventory | Read-only adapters scan all 9 harnesses' on-disk state into one normalized model: sessions, memories, skills, plugins, MCP servers, agents, credential key names, usage. |
🧠 Collective memory | Every harness's native memory imported, deduped, and rendered as an interactive memory graph. Two-way: |
🔐 MCP gateway + vault | One MCP server every harness registers with a single line. Any agent in any harness gets |
🎯 Missions | A shared task board. Write a briefing once, dispatch to any harness with a headless CLI ( |
Quick start
Requires Node ≥ 23.6 (BirdEye runs TypeScript natively — no build step for the daemon).
git clone https://github.com/zanni098/BirdEye.git
cd BirdEye
npm install
npm run build # builds the dashboard (one-time, ~1s)
npm run demo # → http://127.0.0.1:4477 with rich demo data
# or scan YOUR machine:
npm run upThen wire your harnesses into the hub:
node packages/cli/bin/birdeye.ts scan # inventory every harness
node packages/cli/bin/birdeye.ts register claude-code # add the BirdEye MCP gateway (with backup)
node packages/cli/bin/birdeye.ts register codex # ...same one-liner for each harness
node packages/cli/bin/birdeye.ts vault set GITHUB_TOKEN ghp_yourtoken
node packages/cli/bin/birdeye.ts memory sync-back # push shared memory into every context fileThe dashboard
Seven views, live over WebSocket:
Aviary — fleet overview: one card per harness with health, counts, warnings, dispatch capability.
Memory Graph — the shared brain, drawn: memories colored by source harness, tags, auto-extracted entities,
[[links]]. Drag, zoom, filter, click any node for full detail.Missions — task board with live dispatch streaming.
Sessions — cross-harness session timeline with filters.
Arsenal — skills / plugins / MCP matrix: see instantly that
githubMCP is configured 4 separate times.Access — which harness holds which credential keys (names only), with duplication flagged — plus the encrypted vault.
Usage — tokens and cost per harness. Honest: harnesses that don't log usage show unknown, never invented numbers.
Harness support
Harness | Sessions | Memories | Skills | Plugins | MCP | Usage | Dispatch |
Claude Code | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ⚡ |
Claude Desktop | ◐ | — | — | ✅ | ✅ | — | manual |
Codex | ✅ | ✅ | — | — | ✅ | ✅ | ⚡ |
opencode | ✅ | ✅ | ✅ | ✅ | ✅ | ◐ | ⚡ |
OpenClaw | ✅ | ✅ | ✅ | ✅ | ◐ | — | manual |
Zcode | ◐ | ◐ | ✅ | ✅ | — | — | manual |
Gemini CLI | ✅ | ✅ | — | — | ✅ | ◐ | ⚡ |
Cursor | — | ✅ | — | ✅ | ✅ | — | manual |
Continue | ✅ | ✅ | — | — | ✅ | ◐ | manual |
✅ full · ◐ best-effort · — the harness doesn't expose it on disk
The MCP gateway
birdeye register <harness> adds one entry to that harness's MCP config (original backed up first). After that, any agent in any harness can call:
Tool | Does |
| search the unified, deduped memory of all harnesses |
| save a memory every other harness can recall |
| the shared work queue |
| fetch a secret stored once in the encrypted vault |
Security model
The daemon binds
127.0.0.1only. No telemetry, no outbound calls.Adapters are read-only; scanning never touches your harness files.
Credential values are never read from harness configs — key names only. The dashboard never displays secret values;
vault_getflows only over local stdio MCP.Vault: AES-256-GCM, scrypt-derived key (
BIRDEYE_VAULT_PASSPHRASE, or a machine-derived default).The only writers —
register,sync-env,memory sync-back— are explicit commands that create timestamped backups and, for context files, only ever touch the<!-- BIRDEYE:START/END -->block.
How it runs
TypeScript monorepo, zero native dependencies, no compile step for the daemon (Node's native type-stripping). Vite builds the dashboard once.
packages/core types · JSON store · dedup · memory graph · marker sync · vault
packages/adapters 9 harness scanners + registry (failure-isolated)
packages/server Hono daemon: REST + WebSocket + dispatcher + demo mode
packages/mcp the MCP gateway (stdio) + per-harness registration snippets
packages/cli birdeye up · scan · register · vault · sync-env · memory · dispatch
apps/dashboard React + d3-force dashboardnpm test — 62 tests across every package, fixture-driven.
Honest limits (v1)
Usage stats depend on what each harness logs locally — rich for Claude Code/Codex, honest unknown elsewhere.
Cursor/Continue chat data lives in app-internal storage; those adapters are shallower.
Dispatch needs the harness CLI on your PATH.
Cross-harness collaboration is via shared memory + shared task queue; automatic result-chaining between harnesses is on the roadmap.
Roadmap
Result-chaining: harness A's output feeds harness B's next task
Embedding-based memory similarity edges (currently: links, tags, entities)
More adapters — Aider, Windsurf, Cline… (add one — it's ~100 lines)
npx birdeyeas a published npm package
Contributing
Adapters are the sweet spot — one focused file + a fixture + a test. See CONTRIBUTING.md.
MIT © BirdEye contributors
This server cannot be installed
Maintenance
Resources
Unclaimed servers have limited discoverability.
Looking for Admin?
If you are the server author, to access and configure the admin panel.
Related MCP Servers
- AlicenseDqualityDmaintenanceAI-native orchestration layer with 80+ tools for task management, code editing, browser automation, terminal control, and persistent memory across CLI, local MCP, and cloud deployments.Last updated69631ISC
- Alicense-qualityAmaintenanceA local-first control plane for AI agent tools, providing policy enforcement, spend caps, rate limiting, and audit trails for MCP servers.Last updated1Apache 2.0
- Flicense-qualityCmaintenanceA local-first, privacy-preserving centralized memory hub for AI agents — MCP-compatible, zero cloud egress.Last updated2
- Alicense-qualityBmaintenanceMCP server for Hotwired multi-agent workflow orchestration, enabling AI agents to coordinate locally via Unix sockets with tools for protocol, messaging, and task management.Last updated12MIT
Related MCP Connectors
User-owned memory for AI agents, Copilot, Claude, IDEs, CLIs, and chat apps over remote MCP.
Local-first RAG engine with MCP server for AI agent integration.
Private-by-default, local-first memory/context/task orchestrator for MCP apps and agents.
Latest Blog Posts
- Who's Calling? MCP Hosts Are an Identity Blind Spot (And the Spec Knows It)By Om-Shree-0709 on .mcpAgent IdentityOAuth 2.1
- Your AI Chatbot Just Exposed Your CEO's Salary to an InternBy Om-Shree-0709 on .Agent IdentityMCP SecurityOAuth Delegation
- Why MCP Servers Need Execution Sandboxing (And Why Your Current Stack Isn't Enough)By Om-Shree-0709 on .Agentic AiPrompt InjectionWebAssembly
MCP directory API
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/zanni098/BirdEye'
If you have feedback or need assistance with the MCP directory API, please join our Discord server