Deskwise API
A clean REST API for third-party apps. Push contacts, read appointments, and build on top of an operator's workspace — authenticated with a single bearer key and scoped to exactly what you grant.
Authentication
Every request carries an API key as a bearer token. Operators mint keys in Settings → API & licensing. The full key is shown once at creation — store it securely.
curl https://deskwise.co/api/v1/contacts \ -H "Authorization: Bearer dk_live_xxxxxxxx" \ -H "Content-Type: application/json"
Endpoints
| Method | Path | Scope | Description |
|---|---|---|---|
| GET | /api/v1/contacts | contacts:read | List contacts in the workspace. |
| POST | /api/v1/contacts | contacts:write | Create a contact. |
| GET | /api/v1/appointments | appointments:read | List appointments in a window. |
| GET | /api/v1/datasets | datasets:read | List warehoused datasets (name, columns, row counts). |
| POST | /api/v1/datasets | datasets:write | Create a dataset + ingest rows (JSON array or raw CSV) in one call — queryable immediately. |
| PATCH | /api/v1/datasets | datasets:write | Append rows to an existing dataset. |
MCP server — plug Deskwise into any AI
Deskwise is an MCP (Model Context Protocol) server. Point Claude, Claude Code, Claude Desktop, or any MCP-capable client at it and your workspace becomes a set of callable tools — book and reschedule appointments, look up customers, send email and SMS, pull metrics, run link analysis, or route a natural-language request through the AI employee itself with deskwise_ask_assistant.
Keys are minted in Settings → Developers. Every tool is tenant-scoped to the key's workspace — a key can never see another business.
Build on Deskwise — agents, workers, employees
Deskwise is a platform, not just an endpoint. Three ways to build and deploy on it:
- Custom AI employees + orchestration — create specialized agents in the app (Agents → Build a new employee): a name plus an operating prompt. They go live instantly and prompt edits deploy on the next conversation — no code, no release cycle. The lead employee orchestrates them automatically: it delegates matching tasks to your specialists (delegate_to_agent) and acts on their answers in one conversation, one audit trail.
- Background workers— launch long-running jobs on the employee's cloud computer via the MCP tool deskwise_run_task: research, report generation, data analysis with real Python, file deliverables. Poll with deskwise_check_task; results land in the workspace and the org's memory.
- Your own product on top— combine the REST API (contacts, appointments) with the MCP toolset to embed a working employee in your own app or agency stack. Every call is tenant-scoped by key; your client's data never crosses workspaces.
Scopes
Each key carries a space- or comma-separated scope list. Grant the minimum a key needs:
- * — full access to every endpoint.
- contacts:* — any action on contacts.
- contacts:write — exactly that grant.
A request lacking the required scope returns 403 insufficient_scope.
Rate limits
Each key is rate-limited per minute by its plan tier. Over the limit returns 429 with a Retry-After header. Monthly request quotas are metered; usage beyond the quota bills at $0.001 per request.
Pricing
| Plan | Price | Requests / mo | Rate / min | Keys |
|---|---|---|---|---|
| Free | $0/mo | 1,000 / mo | 10 | 1 |
| Solo | $49/mo | 50,000 / mo | 100 | 3 |
| Growth | $199/mo | 250,000 / mo | 500 | 10 |
| Scale | $499/mo | Unlimited (soft cap) | 2,000 | 25 |
Overage on paid tiers is billed at $0.001 per request beyond the monthly quota.
Ready to build?
Create a key in your workspace settings and make your first request in minutes.