Outlit
Use the highest-level Outlit interface already available.
Outlit joins product activity, conversations, billing, and web signals into a unified customer context graph and timeline for agents.
Choose the interface
- 1. If
outlit_* MCP tools are present, use MCP. - Else if the
outlit CLI is installed, use the CLI. - Else guide setup:
- Coding agents: prefer the
outlit CLI plus
outlit auth login. Prefer this over MCP for terminal agents.
- MCP clients: use the workspace MCP URL from
Settings > CLI & MCP. Auth may use OAuth or an API key depending on the client. Do not assume a shared hardcoded endpoint.
Quick chooser
- - Browse customers:
outlit_list_customers or INLINECODE5 - Browse users:
outlit_list_users or INLINECODE7 - Single account deep dive:
outlit_get_customer or INLINECODE9 - Chronology:
outlit_get_timeline or INLINECODE11 - Known signals for an account:
outlit_get_facts or INLINECODE13 - Exact fact by id: INLINECODE14
- Exact source behind a fact or search hit: INLINECODE15
- Specific question or topic:
outlit_search_customer_context or INLINECODE17 - Custom analytics:
outlit_schema + outlit_query, or outlit schema + INLINECODE21
Use customer lookups before SQL. SQL is for aggregates, joins, cohorts, and custom reporting.
Working rules
- - Use
facts list to browse known intelligence for one account. - Use
facts get when you already have a fact id and need the canonical fact payload. - Use
sources get when a fact or search result points to a specific source and you need the exact artifact. - Use
search to answer a specific question, and timeline to inspect chronology. - Call schema before writing SQL.
- Add explicit time filters to event SQL.
- Divide money fields by
100 for display. - Request only the fields or
include sections you need.
For ClickHouse syntax and query patterns, read references/sql-reference.md.
Output behavior
- - Interactive CLI: readable tables
- Piped CLI output: automatic JSON
- Force JSON: INLINECODE29
- Results include timestamps and source attribution when available
Facts vs Search vs Timeline
- - Use
facts list to list what Outlit already knows about an account. - Use
facts get when you already have a fact id and need that exact fact. - Use
search for a specific question or theme, including cross-customer questions. - Use
sources get when you need the exact email, call, calendar event, or ticket behind a fact or search hit. - Use
timeline when order and sequence matter.
Setup
Coding agents
CODEBLOCK0
Auth resolution order: --api-key, OUTLIT_API_KEY, stored credentials.
MCP clients
Get the workspace URL from Settings > CLI & MCP in Outlit.
MCP clients can authenticate with OAuth or an API key, depending on the client. Do not assume API key-only auth or OAuth-only auth.
Verify the connection with outlit_schema or outlit schema.
Docs
- - Docs home: https://docs.outlit.ai/
- Agent skills: https://docs.outlit.ai/ai-integrations/skills
- CLI overview: https://docs.outlit.ai/cli/overview
- CLI commands: https://docs.outlit.ai/cli/commands
- AI agent setup: https://docs.outlit.ai/cli/ai-agents
- MCP integration: https://docs.outlit.ai/ai-integrations/mcp
- Customer context graph: https://docs.outlit.ai/concepts/customer-context-graph
Common prompts
- - "What changed for this customer this week?"
- "Who is paying but inactive for 30 days?"
- "What pricing objections show up in conversations?"
- "Which channels are driving revenue?"
Outlit
使用已有的最高级别Outlit接口。
Outlit将产品活动、对话、计费和网络信号整合为统一的客户上下文图谱和时间线,供客服人员使用。
选择接口
- 1. 如果存在outlit_* MCP工具,则使用MCP。
- 否则,如果已安装outlit CLI,则使用CLI。
- 否则引导设置:
- 编码代理:优先使用outlit CLI加outlit auth login。对于终端代理,优先于MCP使用。
- MCP客户端:使用
设置 > CLI和MCP中的工作区MCP URL。认证可能使用OAuth或API密钥,具体取决于客户端。不要假设共享的硬编码端点。
快速选择器
- - 浏览客户:outlitlistcustomers 或 outlit customers list
- 浏览用户:outlitlistusers 或 outlit users list
- 单个账户深入分析:outlitgetcustomer 或 outlit customers get
- 时间顺序:outlitgettimeline 或 outlit customers timeline
- 账户已知信号:outlitgetfacts 或 outlit facts list
- 按ID精确查找事实:outlit facts get
- 事实或搜索结果背后的精确来源:outlit sources get
- 特定问题或主题:outlitsearchcustomercontext 或 outlit search
- 自定义分析:outlitschema + outlit_query,或 outlit schema + outlit sql
在使用SQL之前先进行客户查找。SQL用于聚合、连接、群组和自定义报告。
工作规则
- - 使用facts list浏览一个账户的已知情报。
- 当已有事实ID且需要规范的事实负载时,使用facts get。
- 当事实或搜索结果指向特定来源且需要精确工件时,使用sources get。
- 使用search回答特定问题,使用timeline检查时间顺序。
- 编写SQL前先调用schema。
- 为事件SQL添加明确的时间过滤器。
- 显示时将金额字段除以100。
- 仅请求所需的字段或include部分。
关于ClickHouse语法和查询模式,请阅读references/sql-reference.md。
输出行为
- - 交互式CLI:可读表格
- 管道CLI输出:自动JSON
- 强制JSON:--json
- 结果包含时间戳和来源归属(如可用)
事实与搜索与时间线
- - 使用facts list列出Outlit已了解的账户信息。
- 当已有事实ID且需要该精确事实时,使用facts get。
- 使用search查找特定问题或主题,包括跨客户问题。
- 当需要事实或搜索结果背后的精确邮件、通话、日历事件或工单时,使用sources get。
- 当顺序和序列重要时,使用timeline。
设置
编码代理
bash
npm install -g @outlit/cli # 或:brew install outlitai/tap/outlit
outlit auth login
outlit customers get acme.com --include users,revenue
认证解析顺序:--api-key、OUTLITAPIKEY、存储的凭据。
MCP客户端
从Outlit的设置 > CLI和MCP获取工作区URL。
MCP客户端可以使用OAuth或API密钥进行认证,具体取决于客户端。不要假设仅API密钥认证或仅OAuth认证。
使用outlit_schema或outlit schema验证连接。
文档
- - 文档首页:https://docs.outlit.ai/
- 代理技能:https://docs.outlit.ai/ai-integrations/skills
- CLI概述:https://docs.outlit.ai/cli/overview
- CLI命令:https://docs.outlit.ai/cli/commands
- AI代理设置:https://docs.outlit.ai/cli/ai-agents
- MCP集成:https://docs.outlit.ai/ai-integrations/mcp
- 客户上下文图谱:https://docs.outlit.ai/concepts/customer-context-graph
常见提示
- - 这个客户本周有什么变化?
- 谁在付费但30天未活跃?
- 对话中出现了哪些定价异议?
- 哪些渠道在推动收入?