MetaHarness: sales vertical

A ready-made multi-agent sales pipeline pod scaffolded onto Claude Code. Spawns three specialized agents — a qualifier (lead scoring), an opener (first-touch outreach), and a closer (negotiation and follow-through) — wired together with shared memory, MCP tooling, and tier-routed model selection. This scaffold gives you a working harness directory you can run, validate, and extend. It does NOT ship a CRM, send live email, or include any contact data — bring your own pipeline source and outbound transport.

Quickstart

npx @metaharness/sales@latest my-bot
cd my-bot && npm install && harness doctor

That gives you a populated harness directory, installed dependencies, and a doctor report telling you which MCP servers / API keys are still missing.

What you get

• agents/qualifier.md — lead-scoring + ICP-fit agent. Tier: haiku (fast, cheap, runs per-lead).
• agents/opener.md — first-touch message drafter, personalization, A/B variants. Tier: sonnet (judgement, tone).
• agents/closer.md — objection handling, negotiation, multi-thread follow-up. Tier: opus (deep reasoning).
• .claude/settings.json — pre-wired hooks (pre-task routing, post-task pattern training, session-start memory restore).
• mcp/ — MCP server stubs for CRM, email, and calendar (stdio transport, swap in your provider).
• prompts/pipeline.md — the orchestration prompt that runs qualifier → opener → closer with SendMessage handoff.
• harness.config.json — tier routing rules, budgets per agent, and shared memory namespace sales-pod.

Advanced

Verify the install is healthy:

$ harness doctor
checking node          ok  v20.11.0
checking claude-code   ok  found in PATH
checking mcp servers   warn  crm: no API key set (CRM_API_KEY)
checking agents        ok  3 agents loaded (qualifier, opener, closer)
checking settings.json ok  hooks valid

Validate the harness manifest before shipping:

$ harness validate
manifest    ok
agents      ok  (3/3 frontmatter valid)
mcp         ok  (3/3 schemas valid)
routing     ok  (tier rules cover all agents)

Run the pod headlessly against one lead, scoped to your scaffold's plugin dir:

$ claude -p --plugin-dir my-bot \
    "Run the sales pipeline on lead: Jane Doe, VP Eng at Acme, 200 employees"
[qualifier→haiku]  ICP fit: 0.82  intent: warm  → handoff to opener
[opener→sonnet]    Drafted 2 variants, selected B (Jane's recent OSS commit)
[closer→opus]      Standing by for reply

FAQ

Does this send real emails? No. The mcp/email stub returns the drafted message and exits. Swap in a real SMTP / Resend / SendGrid MCP server to actually transmit.

Can I run it without API keys? harness doctor will run and the qualifier (haiku) works on offline test fixtures, but the opener and closer call Anthropic — you need ANTHROPIC_API_KEY set for end-to-end runs.

How do I change which model each agent uses? Edit harness.config.json → routing.tiers. Each agent name maps to haiku, sonnet, or opus. The pre-task hook reads this and overrides at dispatch time.

License

MIT. Built on metaharness (https://www.npmjs.com/package/metaharness).

Deep-dive

Full explainer gist: https://gist.github.com/ruvnet/e9cf34f47cf78308196a6e65c4a35322