Integrations — the tool stack¶
AgentCraft artifacts are designed to work with the Hermes Agent .env tool stack. All integrations are optional. Each artifact page lists which .env keys it needs.
What the integration looks like¶
Every skill and flow in AgentCraft reads from the same .env file at ~/.hermes/.env (or wherever your agent loads it from). You set the keys you have; artifacts that need a missing key fail loudly with a clear error.
# Example ~/.hermes/.env
APIFY_TOKEN=apify_api_xxx
DEEPSEEK_API_KEY=sk-xxx
GEMINI_API_KEY=AIzaSy...
EXA_API_KEY=xxx
AIRTABLE_API_KEY=patxxx
RESEND_API_KEY=re_xxx
BEEHIIV_API_KEY=xxx
VERCEL_API_KEY=xxx
GITHUB_API_KEY=ghp_xxx
The 14 tools AgentCraft uses¶
| Tool | Use | Env var | Free? |
|---|---|---|---|
| Apify | LinkedIn scrapers, company search, social data | APIFY_TOKEN |
Pay per actor run |
| DeepSeek | Long-form drafting, enrichment, scoring | DEEPSEEK_API_KEY |
Cheap ($0.14/M input tokens) |
| Gemini | Hero images, section visuals, cover art | GEMINI_API_KEY / GOOGLE_API_KEY |
Free tier, then pay |
| Exa | Research, semantic web search | EXA_API_KEY |
Pay per query |
| Airtable | CRM, lead storage, engagement tracking | AIRTABLE_API_KEY |
Free tier, then $20/mo |
| Resend | Deliverable email to clients | RESEND_API_KEY |
Free tier (3K/mo) |
| Beehiiv | Newsletter drafts + scheduling | BEEHIIV_API_KEY |
Free tier |
| Vercel | Static site deploys | VERCEL_API_KEY |
Free tier (hobby) |
| GitHub | Source repo, public page hosting | GITHUB_API_KEY |
Free |
| OpenRouter | Quality pass / humanize (Claude, GPT) | OPENROUTER_API_KEY |
Pay per token |
| n8n | Workflow automation (optional backend) | n8n instance URL | Self-hosted free |
| Linear | Task tracking, engagement pipeline | Linear OAuth | Free tier |
| Telegram | Cron job delivery, alerts | TELEGRAM_BOT_TOKEN |
Free |
| Firecrawl | Backup web scraper (Exa fallback) | FIRECRAWL_API_KEY |
Pay per page |
What runs offline (no API needed)¶
These artifacts work with zero API keys:
- DEPRO Calculator — pure Python, stdlib only
- API Q1 Audit Checklist — pure Python, stdlib; openpyxl optional for XLSX
- Pareto chart — pure Python
- Fishbone diagram — pure Python + SVG
- DMAIC template — pure Python
- IWCF L4 drill (offline mode) — uses the agent's LLM, but no external APIs
- All scripts — by design, no LLM, no API
Cost discipline¶
The strategy is DeepSeek + Gemini first, OpenAI/Anthropic as the quality pass:
- Drafts and synthesis: DeepSeek (~5x cheaper than GPT-4)
- Images: Gemini Flash (free tier covers most flows)
- Quality pass / humanize: OpenRouter Claude (only on the final 10% of each output)
A typical flow run costs $0.17–$0.55. A full SME audit (7 documents) costs ~$0.54. You're not paying for a SaaS subscription — you're paying per artifact, and the cost is small enough to bill to a client or absorb as overhead.
Adding a new tool¶
When you add a new tool to your .env:
- Update each artifact's
metadata.hermes.requires_envlist in itsSKILL.mdfrontmatter - Add the tool to the
tools_availablelist in thecontext.domainblock - The next CI run will pick it up; the public site integrates page will show the new key automatically
Why this list?¶
These 14 tools are the ones that: - Are present in 90%+ of the underlying 240+ Hermes skills - Have generous free tiers - Are the right cost-per-call for production flows - Don't overlap (each one has a clear, non-overlapping job)
If a new tool displaces one of these on cost or capability, the maintainer updates this list. If a tool has a clear niche and isn't on the list (e.g., Perplexity, Tavily, Browserbase), add it with a one-line use-case.