Give external agents the skills to operate your CRM
DripPulse provides skills packs, OpenAPI contracts, and structured endpoints so agent systems can qualify leads, manage records, run workflows, and act on CRM data safely.
- Works with Claude, OpenClaw, and custom agents
- Built for real CRM operations, not just chat-based suggestions
- Governed access patterns, auth rules, and operational safety guidance included
This page is a first-class access pattern—not a side feature.
Downloads
Canonical artifacts live under /docs/skills/… (static files, not the SPA). If GET returns HTML (often the SPA shell), fix reverse-proxy routing so /docs/skills/ is served as files — see scripts/nginx/drippulse.io.site.conf.template. We do not rely on redirects for these assets; integrators should use the paths below verbatim.
Automated clients: Request Accept: text/markdown for the skills file and Accept: application/json for OpenAPI. Verify the response Content-Type before caching (expect text/markdown / charset=utf-8 or application/json). Pin the v1 filename in your config; bump only when you intentionally adopt a new major skills revision.
Quick start (human in the loop)
- Sign up and create an API key in DripPulse.
- Provide the agent:
DRIPPULSE_BASE_URL(defaulthttps://drippulse.io) andDRIPPULSE_API_KEY. - Agent fetches
drippulse-agent-skills-v1.md(or you paste it into the system prompt), then calls the API using Bearer auth. - Read External agents guide for documentation URLs, what “sessions” means for customers, enrollments vs tests vs
execute, and GUI parity.
Auth contract (agents must follow)
Authorization: Bearer <DRIPPULSE_API_KEY> Accept: application/json Content-Type: application/json
The token determines the organization (tenant). Agents must not attempt to send organization_id to impersonate other tenants.
Minimal “tool” requirement
Any agent framework that can make HTTP requests can drive DripPulse. You only need a single generic tool like:
http.request(method, url, headers, body).
Golden paths (what agents can do immediately)
| Goal | API path(s) | Notes |
|---|---|---|
| Deploy an agent from an out-of-the-box template | /api/v1/agent_templates, /api/v1/agents/spawn, /api/v1/agents/:id/execute |
List templates, spawn an agent with type=template, then execute with an idempotency_key. |
| Create a custom template | /api/v1/agent_templates, /api/v1/agent_templates/:id |
Create/update a prompt-driven template; admins can adjust execution policy. |
| Projects (organize workflows & agents) | /api/v1/projects |
Create/list/update projects, then attach agents/workflows via your orchestration. |
| Webhooks (send events out of DripPulse) | /api/v1/webhooks, /api/v1/webhooks/catalog |
Subscribe to events; verify deliveries using the returned secret (HMAC signature). |
| Create / move opportunities in Kanban | /api/v1/crm/opportunity_pipelines, /api/v1/crm/opportunities |
Fetch pipelines to obtain pipeline_stage_id, then POST/PATCH opportunities. |
| Reorder cards within a stage | /api/v1/crm/opportunities/reorder |
Pass pipeline_stage_id and ordered public_id or UUID list. |
| Maintain accounts / contacts / leads / activities | /api/v1/crm/accounts, /contacts, /leads, /activities |
Standard CRUD. Delete is soft-delete where implemented. |
| Export/import CRM bundles | /api/v1/crm/export, /api/v1/crm/import |
Portability snapshots for migrations/backups. |
| Create/update workflows | /api/v1/workflows |
Use OpenAPI schema + examples; validate before apply. |
| Start live sequences & inspect per-contact runs | POST .../workflows/:id/enrollments/bulk, GET .../enrollments, GET .../enrollments/:id |
List includes step_funnel; dashboard Enrollments tab matches these APIs. |
Claude / OpenClaw guidance
OpenClaw is used inside DripPulse for agent deployment and per-step inference where applicable. Customer-facing automation does not call the OpenClaw gateway directly: use /api/v1 with your org API key. Stateful LLM “sessions” on the gateway are an operator concern; your durable handles are workflow enrollment ids and agent execution records.
Customer integration boundary: External agents guide — durable state vs other “session” meanings. Implementation details: docs/OPENCLAW_INTEGRATION.md in the repo.
What the skills pack contains
- API base URL + auth rules
- Canonical documentation URLs (how agents discover behavior)
- Sessions vs enrollments vs agent executions (customer integration boundary)
- Golden-path sequences for CRM and workflows
- Enrollment list/detail +
step_funneland dashboard alignment - Error handling (401/403/422/429) + retry/backoff rules
- Safety policies (don’t exfiltrate tokens; avoid destructive bulk edits)
- Copy-paste code snippets
Next: read API reference and CRM import & export.