connect
Governed personal world model and memory for your AI agent. Pair once, connect over MCP.
我該用這個嗎
品質與安全性
發現項目（3）
- LOWTool 'odei_connect_status' description lacks action verb在 odei_connect_status 中
- LOWTool 'odei_connect_plan' description lacks action verb在 odei_connect_plan 中
- LOWTool 'odei_connect_talk' description lacks action verb在 odei_connect_talk 中
根據工具定義與協定合規性的自動化分析。
上下文成本
這是每次將伺服器的工具載入模型上下文時所消耗的約略 token 數量。數量越高，可用於其他工作的注意力就越少。
安裝
一鍵安裝
將以下內容加入你的 `claude_desktop_config.json` 檔案：
{
"mcpServers": {
"connect": {
"url": "https://api.odei.ai/mcp"
}
}
}遠端端點
https://api.odei.ai/mcpstreamable-http它能做什麼
工具清單
工具（7）
⚪odei_connect_handshake(pairing_code, agent)
Verify an app.odei.ai pairing code and open an ODEI Connect session.
輸入結構描述
{
"type": "object",
"properties": {
"pairing_code": {
"type": "string",
"description": "Fresh ODEI pairing code from app.odei.ai/profile."
},
"agent": {
"type": "string",
"description": "Optional local AI operator name."
}
},
"required": [
"pairing_code"
],
"additionalProperties": false
}⚪odei_connect_status(session_token, os)
Return current ODEI Connect ceremony progress.
輸入結構描述
{
"type": "object",
"properties": {
"session_token": {
"type": "string"
},
"os": {
"type": "string",
"enum": [
"macos",
"linux",
"windows"
]
}
},
"required": [
"session_token"
],
"additionalProperties": false
}⚪odei_connect_plan(session_token, os)
Return the ordered local ODEI install plan and reporting contract.
輸入結構描述
{
"type": "object",
"properties": {
"session_token": {
"type": "string"
},
"os": {
"type": "string",
"enum": [
"macos",
"linux",
"windows"
]
}
},
"required": [
"session_token"
],
"additionalProperties": false
}⚪odei_connect_report_step(session_token, step, evidence, notes, os)
Record evidence that a Connect ceremony step has completed.
輸入結構描述
{
"type": "object",
"properties": {
"session_token": {
"type": "string"
},
"step": {
"type": "string",
"enum": [
"workspace_created",
"mcp_attached",
"graph_bootstrap_completed",
"first_sync_completed",
"verification_completed"
]
},
"evidence": {
"type": "string"
},
"notes": {
"type": "string"
},
"os": {
"type": "string",
"enum": [
"macos",
"linux",
"windows"
]
}
},
"required": [
"session_token",
"step",
"evidence"
],
"additionalProperties": false
}⚪odei_connect_talk(session_token, message, os)
Answer from ODEI Connect session state without invoking a model API.
輸入結構描述
{
"type": "object",
"properties": {
"session_token": {
"type": "string"
},
"message": {
"type": "string"
},
"os": {
"type": "string",
"enum": [
"macos",
"linux",
"windows"
]
}
},
"required": [
"session_token"
],
"additionalProperties": false
}🟡wm_remember(fact, kind, session_token)
Write one fact to your operator memory in ODEI — the World Model grows from it. Server-side (ODEI-hosted) memory for the paired operator; not the local world-model capsule.
輸入結構描述
{
"type": "object",
"properties": {
"fact": {
"type": "string",
"description": "The fact to remember (max 2000 chars)."
},
"kind": {
"type": "string",
"description": "Optional kind tag, e.g. preference, project, person. Defaults to note."
},
"session_token": {
"type": "string"
}
},
"required": [
"fact"
],
"additionalProperties": false
}🟢wm_recall(query, limit, session_token)
Read facts back from your ODEI-hosted operator memory. Simple substring + recency match (v1) — no semantic search yet.
輸入結構描述
{
"type": "object",
"properties": {
"query": {
"type": "string",
"description": "Optional substring filter over fact text and kind."
},
"limit": {
"type": "number",
"description": "Max facts to return (default 10, cap 50)."
},
"session_token": {
"type": "string"
}
},
"additionalProperties": false
}社群
證據