菜单
Your work knowledge agent. Use Glean chat to answer any question about the user's company, accounts, colleagues, meetings, documents, or work history. Glean synthesizes across 100+ enterprise apps and always cites sources.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Get a pre-meeting briefing for any account. Usage: /brief {account name}. Returns checklist progress, hours remaining, last interaction, missed comms, and open to-dos.
Design and generate Glean Agent specifications. Use when creating new agents, speccing out agent requirements, or generating JSON for Agent Builder import. Triggers on: 'create an agent', 'design an agent', 'agent spec', 'build an agent for [use case]'.
Guidance for effective browser automation with dev-browser plugin. Use for testing local development, verifying UI changes, debugging visual issues, and automating browser tasks.
Design and evaluate context compression strategies for long-running agent sessions. Use when agents exhaust memory, need to summarize conversation history, or when optimizing tokens-per-task rather than tokens-per-request.
A comprehensive collection of Agent Skills for context engineering, multi-agent architectures, and production agent systems. Use when building, optimizing, or debugging agent systems that require effective context management.
Design tools that agents can use effectively, including when to reduce tool complexity. Use when creating, optimizing, or reducing agent tool sets.
| name | glean-mcp |
| description | Your work knowledge agent. Use Glean chat to answer any question about the user's company, accounts, colleagues, meetings, documents, or work history. Glean synthesizes across 100+ enterprise apps and always cites sources. |
Glean is an AI agent with deep context about the user's work — think of it as an oracle for enterprise knowledge. When you're stuck, need background, or want to understand something about the user's company, accounts, colleagues, or work history — ask Glean.
Glean has indexed the user's entire work context:
Glean synthesizes across all these sources. It doesn't just search — it thinks and answers.
chatchat(message="your question here")
chat(message="follow-up question", context=["previous response"])
Glean is identity-aware. It knows the authenticated user automatically:
# These just work — no need to specify the user's name or email
chat(message="What am I working on?")
chat(message="Who is my manager?")
chat(message="What meetings do I have today?")
chat(message="What did I discuss with Jane last week?")
Account & Customer Research:
chat(message="Give me an account overview for MongoDB")
chat(message="Who are the key contacts at Tenstorrent?")
chat(message="What's the deal status for Sports Facilities Advisory?")
chat(message="What use cases is Ratio Therapeutics exploring?")
People & Org Questions:
chat(message="Who works on the agent builder team?")
chat(message="Who should I talk to about MCP integrations?")
chat(message="What's Josh Rutberg's background?")
Process & Policy:
chat(message="How do I escalate a support ticket?")
chat(message="What's the onboarding process for new accounts?")
chat(message="How does the AIOM role differ from CSM?")
Historical Context:
chat(message="What happened in my last meeting with Adam Fowler?")
chat(message="What was decided about the OCR issue at SFC?")
chat(message="What's the history of the MongoDB account?")
Synthesis & Strategy:
chat(message="What are the common blockers for agent adoption?")
chat(message="What patterns do successful agent deployments share?")
chat(message="How do other AIOMs handle high-touch accounts?")
Under the hood, the Glean agent has access to specialized tools. You don't invoke these directly — Glean decides when to use them:
| Tool | What It Does |
|---|---|
| Search | Finds documents across all indexed sources |
| People Lookup | Queries the employee directory and org structure |
| Email Search | Searches Gmail with filters (from, to, labels) |
| Calendar Lookup | Finds meetings and calendar events |
| Document Reader | Retrieves full content from URLs |
| Code Search | Searches internal repositories |
| Activity Tracker | Shows what the user worked on recently |
Glean orchestrates these automatically based on your question.
Always start with chat. If you need more detail, ask follow-up questions:
# Start broad
chat(message="What's the status of the Tenstorrent account?")
# Then drill down
chat(message="What specific use cases are they exploring?",
context=["previous response about Tenstorrent"])
Glean is smart, but specificity helps:
Less effective: "Tell me about MongoDB"
More effective: "What are the current active projects with MongoDB and who are the key stakeholders?"
Use the context parameter for follow-ups:
response1 = chat(message="What meetings do I have with Ratio Therapeutics?")
response2 = chat(
message="What should I prepare for the next one?",
context=[response1]
)
Every response includes source links. These are real — use them to verify or dive deeper.
| Situation | Ask Glean |
|---|---|
| Starting work on an account | "Account overview for X" |
| Preparing for a meeting | "Prep me for my meeting with X" |
| Researching a person | "What do I know about X?" |
| Understanding a project | "What's the status of X?" |
| Finding an expert | "Who knows about X?" |
| Recalling a decision | "What was decided about X?" |
| Writing a summary | "Summarize my activity on X" |
| Investigating an issue | "What's the history of X issue?" |
| Need | Use Instead |
|---|---|
| Public/external information | Web search |
| Local project files | Read tool |
| Info already in conversation | Reference it directly |
| Real-time data | Glean indexes periodically |
| Speculation/opinion | Your own reasoning |
Indexing lag: New documents may take minutes to hours to appear.
Permission-scoped: Glean only sees what the user has access to. If results seem sparse, the user may lack permissions.
Structured data: Returns markdown/snippets, not raw CSVs. For full spreadsheet analysis, have the user upload the file.
External companies: Glean knows about the user's company's interactions with external companies (emails, meetings, CRM data) but doesn't have access to their internal systems.
# Before a customer call
chat(message="Prep me for my call with Adam Fowler at Sports Facilities. Include recent context, open issues, and what we discussed last time.")
# New account handoff
chat(message="Give me a full briefing on the Golden Gate Bridge account — adoption status, key contacts, risk factors, and what the previous AIOM was working on.")
# Debugging an issue
chat(message="What do we know about the agent error Adam Fowler reported? Include request IDs and any support ticket context.")
# Start of week
chat(message="What are my priorities this week based on my calendar, recent activity, and outstanding tasks?")
# Before a 1:1
chat(message="What should I know before my 1:1 with Josh Rutberg? Include recent discussions and any items I should bring up.")
Glean exists to reduce cognitive load. Instead of:
Just ask Glean. It handles the complexity. You get the answer.
This is the "second brain" pattern — an AI agent with deep context about your work, always available to consult when you need to understand, remember, or decide.
When in doubt, ask Glean.