LLM call recording

Trace every LLM call across 9 providers — cost, latency, and errors, one by one.

Setup is one line pasted into your coding agent, or one line of SDK.

Recent calls (example)

ModelCostStatus
gpt-5.5<$0.01Success

Prompt

Reply politely and concisely to a customer who says they cannot log in.

Completion

Sure, let me check the situation. Could you try resetting your password from the link below?

Servicesupport-bot
Useruser_a12f
Sessionsess_8f12
TokensInput 280 / Output 96
Latency1.14 s

9 providers supported

Wrap the OpenAI / Anthropic / Gemini / Mistral SDK clients — Grok, Kimi, DeepSeek, Qwen, and Muse Spark work through the OpenAI-compatible API. Argosvix records the rest automatically.

OpenAIAnthropicGeminiMistralGrokMoonshot KimiDeepSeekAlibaba QwenMeta Muse Spark

Per-call detail you can trace

Token counts, USD cost, latency, and errors — per call, in one view. Full prompt/completion storage is an opt-in Pro+ option (all supported providers).

Example
Call detailAnthropicclaude-sonnet-5

Cost (USD)

$0.0049

Total tokens

1,763

Input

1,590

Output

173

Cache hit

32%

Latency

3963 ms

Status

success

Error

Full prompt/response storage is a Pro+ opt-in.

Dashboard, MCP & REST

Reach the same data from the web UI, from Claude Desktop / Cursor / Codex CLI, or from curl.

UIdashboard.argosvix.com
MCPquery_calls
RESTGET /v1/query/calls