🎯 events-agent

An AI events research agent for partnerships and growth teams. Fork it, customise it for your company, run it daily.

---

⚠️ Independent community project. This is an open-source side project, not affiliated with, endorsed by, or supported by Anthropic. "Claude" and "Claude Code" are products of Anthropic, used here per their standard terms of service. You bring your own Claude subscription or API key — this template doesn't host or proxy anything.

---

⚡ Try it in 3 minutes — no setup required

Just want to see what it does? Open TRY_IT.md →

Copy a prompt, paste it into Claude.ai, and you have a working events research agent in your next conversation. No fork, no terminal, no GitHub knowledge needed.

When you're ready for the persistent version, read on.

---

Why this exists

If you work in partnerships, BD, or growth marketing, you probably lose 5-15 hours a week to events research. The same shape of work, every week:

• Pull up a website to check the audience
• Cross-reference the sponsor list against your competitors
• Look up speaker submissions and CFP deadlines
• Calculate flights, hotels, ticket costs from your home base
• Estimate pipeline impact based on ICP density
• Compare it against the four other events the same week
• Write up a verdict for your boss / CEO

This template turns that into a 30-minute review of an agent's daily output. Same quality of decisions. 95% less grinding.

The trick isn't using AI. It's giving AI structured, persistent context — your ICP, your competitors, your priorities — so every session doesn't start from zero.

What it does

Each day, the agent:

1. Searches for new events matching your ICP (verticals, target accounts, geographies)
2. Cross-references sponsor lists for your competitors (red flags) and clear lanes (green flags)
3. Scores each event using a consistent verdict framework (GO / WATCH / SKIP / SPONSOR / SPEAKER-FIRST)
4. Estimates pipeline (qualified conversations → demos → opportunities) and travel costs
5. Surfaces CFP deadlines, ticket discount expiries, and startup tier opportunities
6. Outputs a structured Markdown report you can paste into Notion, Slack, or your weekly partnerships review

Three setup levels depending on how much automation you want:

• Level 1 — Claude Project (10-min setup, free with Claude Pro, manual trigger)
• Level 2 — Claude Code automation (~30-min setup, daily cron, ~$1-3/day in API credits)
• Level 3 — Custom MCP server (1-week build, full team access via any Claude chat)

🚀 Quick start (Level 1, 10 minutes)

Step 1: Fork + clone

gh repo fork lilianobennett-ai/events-agent --clone
cd events-agent

Step 2: Copy an example config and customise

# Pick the closest fit to your business model
cp -r examples/saas-feedback-tool/config/* config/
# OR
cp -r examples/dev-tools/config/* config/

Open config/ and edit five files:

File	What it contains
01-company.md	Who you are, what you sell, what makes you different
02-icp.md	Tier 1/2/3 ideal customer profiles + named target accounts
03-verticals.md	Industries you sell into, ranked by 2026 priority
04-competitors.md	Tier 1 (hostile direct), Tier 2 (legacy / pivoting), Tier 3 (adjacent)
05-priorities.md	Budget posture, geographic preferences, speaker-first vs sponsor-first

Be concrete and ranked, not aspirational. The agent's quality is proportional to your config specificity. Vague configs produce vague recommendations.

Step 3: Generate your system prompt

./scripts/build-prompt.sh

This concatenates your 5 config files with templates/system-prompt-base.md and outputs your-system-prompt.md.

Step 4: Set up the Claude Project

1. Go to claude.ai/projects → "Create Project"
2. Name it: "[Your Company] Events Research Agent"
3. Paste contents of your-system-prompt.md into "Custom Instructions"
4. (Optional) Upload past event analyses as Project knowledge so the agent doesn't re-recommend events you've already evaluated
5. Save

Step 5: Use it daily

Open the project and ask:

Find me 10 new events for Q4 2026 / Q1 2027 that match our criteria,
sorted by verdict.

Or:

Deep dive on [event name] — agenda, sponsor list, ICP fit, competitor presence.

Or:

What CFP deadlines are closing this month?

That's it.

🤖 Level 2 — Daily automation

Want it to run on a schedule, write reports to disk, post to Slack? See claude-code/README.md for the cron-based setup.

Cost: ~$1-3/day. Tested with Claude Code v2.1.111+ on macOS and Linux.

🏗️ Level 3 — MCP server (full team access)

If you want anyone on your team to query the agent from any Claude conversation, build a custom MCP server. PRs welcome to add a reference implementation.

🤝 Contributing

This is a community-maintained template. PRs welcome for:

• New industry-specific configs (we have B2B SaaS + dev tools — add fintech, cybersecurity, adtech, hardware)
• New event source integrations (Sessionize, Luma, AAIF, regional newsletters)
• Improvements to the verdict framework
• Bug fixes in the Claude Code automation

See CONTRIBUTING.md.

📜 License

MIT. Use it, fork it, customise it. Build on it commercially if you want — just keep the original copyright in the LICENSE file.

🙋 FAQ

Q: Is this affiliated with Anthropic / Claude? No. Independent community project. Not affiliated with, endorsed by, or supported by Anthropic. You bring your own Claude Pro subscription (Level 1) or Anthropic API access (Level 2/3).

Q: Will this work with GPT / Gemini / other models? The system prompt is largely model-agnostic. Tested on Claude Opus 4.7 and Claude Sonnet 4.6 — both work well. Other models should work with prompt tweaks but I haven't tested them. PRs with adaptations welcome.

Q: How is this different from manually using ChatGPT? Three things: (1) structured system prompt encodes your ICP / competitors / verdict framework so output is consistent across sessions, (2) configs are explicit so the agent never re-recommends events you've already evaluated, (3) the workflow is repeatable and shareable across your team.

Q: I'm not technical — can I still use this? Yes. Level 1 requires zero code. You edit 5 markdown files in any text editor, copy-paste into Claude.ai, done. Level 2 needs basic command-line comfort.

Q: What about my data privacy? Level 1 sends your config to Claude's servers (subject to Anthropic's privacy policy). Level 2/3 keeps configs local and only sends queries via API.

Q: How much does it cost?

• Level 1: Free with Claude Pro (~$20/month)
• Level 2: ~$30-90/month in API credits depending on run depth
• Level 3: Same as Level 2 plus your hosting costs

Q: Can I use this commercially? Yes — MIT license. Use inside your company, modify it, integrate into commercial products. Just keep the LICENSE file's copyright notice.

---

Built as a personal side project by Lilian Bennett. ⭐ if it saves you time. PRs if you want to make it better for everyone.