API Documentation|

Your API key starts with sch_prod_. Find it in your dashboard after activation.

cURL

curl -X POST https://sch-provision-api.onrender.com/v1/agents/YOUR_AGENT_ID/chat \
  -H "X-API-Key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"message": "Hello, what can you help me with?"}'

Python (httpx)

import httpx

r = httpx.post(
    "https://sch-provision-api.onrender.com/v1/agents/YOUR_AGENT_ID/chat",
    headers={"X-API-Key": "YOUR_API_KEY"},
    json={"message": "Hello, what can you help me with?"},
)
print(r.json()["response"])

JavaScript (fetch)

const res = await fetch(
  "https://sch-provision-api.onrender.com/v1/agents/YOUR_AGENT_ID/chat",
  {
    method: "POST",
    headers: { "X-API-Key": "YOUR_API_KEY", "Content-Type": "application/json" },
    body: JSON.stringify({ message: "Hello!" }),
  }
);
const data = await res.json();
console.log(data.response);

Pass your API key in every request via the X-API-Key header:

X-API-Key: sch_prod_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

All errors return JSON: {"success": false, "error": "description"}

{
  "status": "ok",
  "version": "1.2.0"
}

{
  "response":         "Hello! I can help you with...",
  "tokens_used":      312,
  "tokens_remaining": 499688,
  "latency_ms":       1204
}

curl -X POST https://sch-provision-api.onrender.com/v1/agents/AGENT_ID/chat \
  -H "X-API-Key: YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{"message":"Summarise my last 5 support tickets","session_id":"sess_abc123"}'

{
  "plan":             "starter",
  "tokens_allocated": 500000,
  "tokens_used":      12500,
  "tokens_remaining": 487500,
  "pct_used":         2.5,
  "expires_at":       "2026-07-01T00:00:00+00:00",
  "status":           "active",
  "recent_calls": [
    {
      "tokens":          312,
      "latency_ms":      1204,
      "created_at":      "2026-05-10T09:30:00Z",
      "message_preview": "Hello, what can you..."
    }
  ]
}

{
  "agent_id":         "a1b2c3d4e5f6a7b8",
  "api_key":          "sch_prod_...",
  "endpoint_url":     "https://sch-provision-api.onrender.com/v1/agents/a1b2c3d4e5f6a7b8/chat",
  "plan":             "starter",
  "tokens_allocated": 500000,
  "expires_at":       "2026-06-18T00:00:00+00:00",
  "status":           "pending_payment"
}

{ "success": true, "message": "OTP sent" }

{
  "order_id": "order_PxyzABC123",
  "amount":   1900,
  "currency": "USD",
  "key_id":   "rzp_live_..."
}

{
  "success": true,
  "message": "Payment verified. Agent is now active.",
  "status":  "active"
}

{
  "success":   true,
  "refund_id": "rfnd_PxyzABC123",
  "amount":    99900,
  "status":    "processed"
}

curl https://sch-provision-api.onrender.com/api/invoice/AGENT_ID \
  -H "X-API-Key: YOUR_KEY" \
  --output invoice.pdf

Every active workflow on your agent is automatically exposed as an MCP tool. Connect Claude Desktop, Claude Code or Cursor to your agent and they can list and run your workflows directly — same run history, approval gates and token accounting as running them from the dashboard. Workflows in draft status stay private; flipping one to active is what publishes it.

Endpoint & auth

POST https://sch-provision-api.onrender.com/mcp
Authorization: Bearer sch_prod_YOUR_API_KEY   (or X-API-Key header)

Stateless Streamable HTTP JSON-RPC. Your agent API key is the credential — the same active/expiry gate as the chat endpoint applies.

Claude Code (one command)

claude mcp add --transport http agonix \
  https://sch-provision-api.onrender.com/mcp \
  --header "Authorization: Bearer sch_prod_YOUR_API_KEY"

Claude Desktop / Cursor (config JSON)

{
  "mcpServers": {
    "agonix": {
      "type": "http",
      "url": "https://sch-provision-api.onrender.com/mcp",
      "headers": { "Authorization": "Bearer sch_prod_YOUR_API_KEY" }
    }
  }
}

What the client sees

Each active workflow appears as a tool named wf_<id>_<name> accepting a free-form JSON object as the workflow's trigger input. A call returns the run id, success flag, tokens used and node outputs. If the workflow hits an approval node, the call returns paused_for_approval: true and the run waits for your approve/reject in the dashboard — a human stays in the loop even when another AI triggered the run.

Which model runs your agent depends on its plan. Free tiers route to a fast open model with automatic fallbacks so a single provider outage never takes your agent down; enterprise routes to Anthropic Claude.

Token accounting. Every call sums input + output (+ cache) tokens against your monthly allocation. Past the allocation you may continue into overage at $0.002 / 1,000 tokens up to a hard ceiling of 2× your quota, after which the agent returns 402 until you upgrade or the quota renews. Usage is visible in real time at GET /usage.

Data handling. Agent input and output pass a guardrail layer: card numbers, national IDs and API keys/tokens are redacted before any model sees them and again before a response leaves the API; emails and phone numbers are stripped from the stored usage-log preview. Prompt-injection attempts are logged, not silently executed. Connector and MCP credentials are encrypted at rest (Fernet). Conversation memory is stored per-agent (pgvector) to give continuity across runs and is never shared across tenants.

Connectors. 30 apps have live actions today (CRM, payments, comms, productivity, e-commerce) plus a universal HTTP action for any public API, and external MCP servers can be registered to add their tools. Browse the full catalog at /connectors/actions.

Tokens count input + output + cache tokens. Quotas reset on renewal and do not roll over. You receive an email alert at 80% usage and again at 100% (when the agent pauses).

Limits are per IP. When exceeded you receive 429 Too Many Requests with a Retry-After header in seconds.