lighthouse

name

lighthouse

description

Use Lighthouse as an end user — daily research briefs with BYOK AI, topic selection, custom topic builder, and delivery channels. Connect via WebMCP tools on the live dashboard or guide the user through the web UI. Use when the user wants research digests, wants to run or schedule briefs, connect an agent to Lighthouse, or operate the demo at demo-beta-topaz.vercel.app — not when developing or contributing to the lighthouse-kestra repository.

Lighthouse (usage)

Lighthouse answers: “What moved today in the spaces I care about?”
Live app: https://demo-beta-topaz.vercel.app
Your keys: BYOK — LLM API keys stay in the user’s browser, never on Lighthouse servers.

Install this skill

npx skills add guglxni/lighthouse-kestra --skill lighthouse -a claude-code -a cursor -a codex -a gemini-cli -y

What you can do for the user

Goal	How
Sample a brief now	Dashboard → pick Topics → Generate a sample brief
Track a new niche	Build your brief → describe scope → topic auto-saves and activates
Daily delivery	Settings → webhooks / AgentMail / Slack or Notion OAuth
Agent drives the UI	WebMCP tools on the signed-in dashboard (see below)

User flow (web UI)

1. Sign in at /login
2. Settings (/settings) — add LLM API key, base URL, model (required for briefs)
3. Dashboard (/dashboard)
   • Topics — click a card (preset or custom); this sets the active topic
   • Generate a sample brief — dropdown mirrors the selected topic; run with a question
   • Build your brief — plain-English scope + simple schedule (how often + time UTC); YAML validates and plugs in automatically
4. Settings again — optional Slack/Discord/Telegram/Notion/email delivery

Custom topics: saved to the user’s account, selected automatically after creation, and injected into sample-brief prompts (sources + scope from their Kestra YAML).

WebMCP — connect your agent to Lighthouse

Open dashboard or settings while signed in. Lighthouse registers WebMCP tools with write-only secrets (masked status only).

Secure setup (authsome + WebMCP)

1. lighthouse-authsome-setup-guide — CLI steps per provider
2. User runs authsome login slack / notion / stores OpenAI key locally
3. Import without echoing secrets back:
   • lighthouse-configure-llm — BYOK in browser
   • lighthouse-import-authsome-token — Slack/Notion server-side (from authsome get … --field access_token)
   • lighthouse-start-oauth — browser OAuth alternative
4. lighthouse-get-config-status — verify (masked fingerprints only)
5. lighthouse-configure-delivery / lighthouse-configure-agentmail
6. Briefs: lighthouse-list-topics → lighthouse-select-topic → lighthouse-run-sample-brief

Tool	Purpose
lighthouse-get-config-status	Masked config snapshot
lighthouse-configure-llm	Store LLM key (write-only)
lighthouse-configure-exa	Optional Exa key
lighthouse-configure-agentmail	AgentMail BYOK + inbox
lighthouse-configure-delivery	Webhooks, Telegram, Notion page, default topic
lighthouse-import-authsome-token	Slack/Notion token from authsome CLI
lighthouse-start-oauth	Open Slack/Notion OAuth tab
lighthouse-authsome-setup-guide	authsome CLI instructions
lighthouse-send-test-email	Test AgentMail wiring
lighthouse-list-topics	Topics + active selection
lighthouse-run-sample-brief	Draft brief (BYOK)
lighthouse-build-topic	Custom topic from brief

Never repeat full API keys or tokens in agent chat after import. Use lighthouse-get-config-status instead.

Spec: https://github.com/webmachinelearning/webmcp · authsome: https://authsome.ai/docs/llms.txt

Schedule (user-facing)

Users pick How often? (every day / weekdays / weekly Monday) and What time? (UTC). No raw cron required.
Examples: every day at 9:00 UTC · weekdays at 8:00 UTC.

Delivery channels

Channel	User setup
Email	AgentMail API key + inbox in Settings — https://docs.agentmail.to/llms.txt
Slack	OAuth connect in Settings — https://authsome.ai/docs/integrations/oauth/slack
Notion	OAuth connect in Settings — https://authsome.ai/docs/integrations/oauth/notion
Discord / Telegram	Webhook or chat id in Settings

Agent checklist

- [ ] User signed in?
- [ ] BYOK configured in Settings?
- [ ] Topic selected (card or lighthouse-select-topic)?
- [ ] For custom scope: build topic first OR use existing custom topic
- [ ] Run sample brief or configure delivery

Preset topics

id	Focus
agentic-eng	AI agents, MCP, coding assistants
solana-zk	Solana + ZK / audits
indie-saas	Bootstrapped SaaS, growth
data-eng-ai	Data engineering, RAG, lakehouse

Additional resources

• WebMCP tool schemas & fallbacks: webmcp.md
• End-user workflows & troubleshooting: using-lighthouse.md

Not in scope for this skill: editing Kestra flows, contributing to the repo, or deploying infrastructure — use the project README for that.