Agent Public API

The Agent Public API lets you call the Tritonix investment agent from your own server-side scripts and backend jobs. It uses the same core execution engine as Tritonix headless agent runs, including built-in research tools and expert subagents, while staying fully separate from front-end chat history. It also does not load user-connected MCP / brokerage tools.

Base Endpoint

POST https://tritonix.ai/api/v1/agent/run

Use the production domain above for live integrations. If you are testing against a local or preview deployment, replace https://tritonix.ai with that environment's base URL.

Authentication

Send your API key in the Authorization header:

Authorization: Bearer txk_your_secret_here
Content-Type: application/json

Request Body

Minimal request

{
  "prompt": "Before the US market open, select 5 stocks for today's watchlist and return valid JSON.",
  "model": "gemini-3.7-flash",
  "response_format": {
    "type": "json_object"
  }
}

Fields

Field	Required	Type	Description
`prompt`	Yes	string	The instruction sent to the agent
`model`	No	string	Tritonix model key. Unknown values fall back to `gemini-3.7-flash`
`response_format`	No	object	Output mode. Supported: `text`, `json_object`, `json_schema`

Supported Models

Pass the model key as a string in the model field.

Current supported keys:

gpt-5.6-terra
gpt-5.6-luna
claude-opus-5
claude-sonnet-5
claude-haiku-4-5
gemini-3.1-pro-preview
gemini-3.7-flash
grok-4.6
grok-4.1-fast-reasoning
deepseek-v4-pro
kimi-k2.6
qwen3.7-plus
glm-5.2
minimax-m3

Notes:

If model is omitted, Tritonix uses gemini-3.7-flash
If an unknown model key is provided, Tritonix also falls back to gemini-3.7-flash

Parameter Reference

`prompt`

Type: string
Required: Yes
Purpose: the full instruction for the agent
Maximum: 800,000 tokens for models with sufficient context. Tritonix automatically uses a lower effective limit for smaller-context models and truncates excess input instead of failing the request.
The response usage object reports submitted_input_tokens, retained_input_tokens, input_token_limit, and input_truncated.

`model`

Type: string
Required: No
Example: "gemini-3.7-flash"

`response_format`

Type: object
Required: No
Default: { "type": "text" }

Supported values:

{ "type": "text" }

{ "type": "json_object" }

{
  "type": "json_schema",
  "schema": {
    "type": "object",
    "properties": {
      "answer": { "type": "string" }
    },
    "required": ["answer"]
  }
}

Response Formats

1. Text output

{
  "prompt": "Summarize NVIDIA in plain English.",
  "response_format": { "type": "text" }
}

Returns plain text in output.

2. Arbitrary JSON object

{
  "prompt": "Return a JSON watchlist with tickers and reasons.",
  "response_format": { "type": "json_object" }
}

Returns parsed JSON in output, without schema enforcement.

3. JSON schema output

{
  "prompt": "Return exactly 3 premarket picks.",
  "response_format": {
    "type": "json_schema",
    "schema": {
      "type": "object",
      "properties": {
        "picks": {
          "type": "array",
          "items": {
            "type": "object",
            "properties": {
              "ticker": { "type": "string" },
              "market": { "type": "string" },
              "reason": { "type": "string" }
            },
            "required": ["ticker", "market", "reason"]
          }
        }
      },
      "required": ["picks"]
    }
  }
}

Returns validated structured output in output.

Example Successful Response

{
  "id": "0fdfce46-6a63-4bf2-a70e-4d7bced1f8f6",
  "model": "gemini-3.7-flash",
  "response_format": "json_object",
  "output": {
    "picks": [
      {
        "ticker": "NVDA",
        "reason": "Earnings momentum remains strong and premarket liquidity is deep."
      }
    ]
  },
  "usage": {
    "prompt_tokens": 4821,
    "completion_tokens": 533,
    "total_tokens": 5354,
    "cache_read_tokens": 0,
    "cache_write_tokens": 0,
    "reasoning_tokens": null,
    "agent_steps": 3,
    "cost_usd": 0.008019,
    "credits_deducted": 3
  }
}

Error Responses

Invalid API key

{
  "error": "invalid_api_key"
}

Status: 401

Missing prompt

{
  "error": "missing_prompt"
}

Status: 400

Insufficient credits

{
  "error": "insufficient_credits",
  "balance": 0
}

Status: 402

Execution failure

{
  "error": "agent_run_failed",
  "message": "..."
}

Status: 500

Credits Behavior

Credits follow Tritonix's existing platform logic:

if credits_balance <= 0 before execution, the request is rejected
if credits_balance > 0 before execution, the request may run
final credits deduction happens after execution

Public API requests are also recorded in the API invocation log for dashboard analytics.

Operational Metadata

Tritonix stores request-level metadata for the public API, including:

request status
model
response format
token usage
Credits deducted
latency
error code

Tritonix does not store the full prompt body or full model response body for this public API.

cURL Example

curl -X POST "https://tritonix.ai/api/v1/agent/run" \
  -H "Authorization: Bearer txk_your_secret_here" \
  -H "Content-Type: application/json; charset=utf-8" \
  -d '{
    "model": "gemini-3.7-flash",
    "prompt": "Before the market opens, return 3 US stock ideas in JSON with ticker, theme, and risk.",
    "response_format": {
      "type": "json_object"
    }
  }'

Windows PowerShell Encoding

When sending non-English prompts from Windows PowerShell, pass the JSON body as UTF-8 and set the content type explicitly:

$headers = @{ Authorization = 'Bearer txk_your_secret_here' }
$body = @{
  model = 'gemini-3.7-flash'
  prompt = '请用一句话介绍一下英伟达最近为什么仍然值得投资者关注。'
  response_format = @{ type = 'text' }
} | ConvertTo-Json -Depth 5

$resp = Invoke-RestMethod `
  -Method Post `
  -Uri 'https://tritonix.ai/api/v1/agent/run' `
  -Headers $headers `
  -ContentType 'application/json; charset=utf-8' `
  -Body ([System.Text.Encoding]::UTF8.GetBytes($body))

$resp.output

Related Docs

API Key Management