@voightxyz/anthropic

Voight observability for the Anthropic SDK. Wrap your Anthropic client and capture every Messages call — prompts, tokens, cache reads, cache creations, tool use, costs, latency, errors — surfaced live in the Voight dashboard.

Same backend and dashboard as @voightxyz/openai. Drop in whichever provider your app uses; events from both land side-by-side under the same agent.

Quick setup with the wizard

If your app already imports @anthropic-ai/sdk, the lowest-friction install is the wizard from the main SDK:

cd your-app
npx -y @voightxyz/sdk init

It detects @anthropic-ai/sdk (and openai if present) in your package.json, prompts for a privacy level + Voight key + agent name, and writes a ready-to-import src/lib/voight.ts with the wrapped client. 30 seconds, zero copy-paste. Full walkthrough at docs.voight.xyz/ai-apps/wizard.

Continue below if you'd rather wire it manually.

Install

npm install @anthropic-ai/sdk @voightxyz/anthropic

Quick start

import Anthropic from '@anthropic-ai/sdk'
import { wrapAnthropic } from '@voightxyz/anthropic'

const client = wrapAnthropic(new Anthropic(), {
  voightApiKey: process.env.VOIGHT_KEY,
  agent: 'my-prod-agent',
})

const response = await client.messages.create({
  model: 'claude-haiku-4-5',
  max_tokens: 1024,
  messages: [{ role: 'user', content: 'Hello' }],
})

That's it — every call is captured automatically. Visit your Voight dashboard to see them in real time.

Tracing & per-user attribution

For production apps where you want to group every LLM call inside one request into one trace, and attribute cost per end-user with one line of code, wrap each request boundary with withTrace:

import Anthropic from '@anthropic-ai/sdk'
import { wrapAnthropic, withTrace, log } from '@voightxyz/anthropic'

const anthropic = wrapAnthropic(new Anthropic(), {
  agent: 'production-chat-api',
  privacy: 'standard',
})

app.post('/api/chat', async (req, res) => {
  await withTrace(
    async () => {
      log('chat request received')
      const reply = await anthropic.messages.create({
        model: 'claude-haiku-4-5',
        max_tokens: 1024,
        messages: [{ role: 'user', content: req.body.prompt }],
      })
      res.json({ reply })
    },
    {
      routeTag: 'POST /api/chat',
      tags: { userId: req.user.id, plan: req.user.plan },
    },
  )
})

Every wrapped LLM call inside the withTrace block automatically inherits the routeTag and tags. The tags.userId field drives per-user spend tracking — the dashboard's Users sub-tab populates with per-customer cost as soon as your first request lands.

The same withTrace exported here also works in @voightxyz/openai — they share an async-context store, so an app calling both providers inside one request gets one trace, not two.

What's captured

| Signal | Where it lands | |---|---| | Model id (with version suffix) | model | | Prompt messages | input.messages | | Response text (aggregated from content[].text blocks) | metadata.responseText | | Token counts (input / output / total) | metadata.tokens | | Cache reads (cache_read_input_tokens) | metadata.tokens.cache_read | | Cache creations (cache_creation_input_tokens) | metadata.tokens.cache_creation | | Tool use (full array) | metadata.toolCalls + toolExecuted | | Streaming flag | metadata.streaming | | Trace grouping (auto UUID or explicit) | metadata.sessionId | | Stop reason | metadata.finishReason | | Latency (ms) | durationMs | | Errors (re-thrown to the caller) | errorMessage + outcome: 'failed' |

Supported endpoints

• client.messages.create — Messages API (non-streaming + streaming, tool use, cache breakpoints)

The wrapper passes everything else through untouched. Bedrock and Vertex clients are on the 0.2.0 roadmap.

Options

| Option | Type | Default | Purpose | | --- | --- | --- | --- | | voightApiKey | string | process.env.VOIGHT_KEY | Your Voight key from the dashboard | | agent | string | process.env.VOIGHT_AGENT → HOSTNAME → 'unknown-agent' | Stable identifier surfaced in the dashboard | | apiBase | string | https://api.voight.xyz | Override for self-hosted deployments | | privacy | 'minimal' \| 'standard' \| 'full' | 'standard' | Capture aggressiveness | | sessionId | string | auto UUID v4 | Trace grouping. Stable across calls of one wrapper instance | | enabled | boolean | true | Kill switch — returns the original client untouched |

Privacy

Three levels apply to prompts, response text, and tool-call arguments. The function name in toolExecuted always survives as a tag (not user content).

| Level | Prompts | Response text | Tool arguments | Tokens / timing / model | | --- | --- | --- | --- | --- | | minimal | dropped | dropped | dropped | kept | | standard (default) | scrubbed | scrubbed | scrubbed | kept | | full | verbatim | verbatim | verbatim | kept |

Standard scrubs 12 patterns: PEM private keys, JWTs, Anthropic / OpenAI / Stripe live / GitHub / AWS / Slack / Voight API keys, emails, E.164 phones, and Luhn-validated credit cards.

See CHANGELOG.md for release notes.

License

Apache 2.0