Display Name
Voight
Category
Hooks
Sub-Category
None
Primary Link
https://github.com/Voightxyz/voight-sdk
Author Name
Voight
Author Link
https://github.com/Voightxyz
License
Apache-2.0
Other License
No response
Description
Real-time observability for Claude Code sessions via the hooks API. Captures every prompt, tool call, token cost, and error to a hosted dashboard with cross-session tracing. Three privacy levels with local PII scrubbing before transmission.
Validate Claims
Production deployments at voight.xyz with active users. 272 unit tests on the SDK (Apache 2.0). Verified end-to-end against real Claude Code sessions emitting events through PreToolUse / PostToolUse / Stop / UserPromptSubmit / PreCompact hooks.
Specific Task(s)
Monitor LLM cost and latency per agent and per session. Detect anomalies (cost spikes, error rate, orphan tool calls). Cross-session search via traceId. Per-tool execution breakdown.
Specific Prompt(s)
After installing via npx -y @voightxyz/sdk setup, start any Claude Code session — events appear live in the dashboard. Example: ask Claude to refactor a file; the dashboard shows the prompt, every tool call, tokens, cost, and the assistant's response.
Additional Comments
Apache 2.0 OSS. Also provides companion App SDKs (@voightxyz/openai, @voightxyz/anthropic, @voightxyz/vercel-ai) for production LLM app monitoring beyond just Claude Code.
Recommendation Checklist
Display Name
Voight
Category
Hooks
Sub-Category
None
Primary Link
https://github.com/Voightxyz/voight-sdk
Author Name
Voight
Author Link
https://github.com/Voightxyz
License
Apache-2.0
Other License
No response
Description
Real-time observability for Claude Code sessions via the hooks API. Captures every prompt, tool call, token cost, and error to a hosted dashboard with cross-session tracing. Three privacy levels with local PII scrubbing before transmission.
Validate Claims
Production deployments at voight.xyz with active users. 272 unit tests on the SDK (Apache 2.0). Verified end-to-end against real Claude Code sessions emitting events through PreToolUse / PostToolUse / Stop / UserPromptSubmit / PreCompact hooks.
Specific Task(s)
Monitor LLM cost and latency per agent and per session. Detect anomalies (cost spikes, error rate, orphan tool calls). Cross-session search via traceId. Per-tool execution breakdown.
Specific Prompt(s)
After installing via
npx -y @voightxyz/sdk setup, start any Claude Code session — events appear live in the dashboard. Example: ask Claude to refactor a file; the dashboard shows the prompt, every tool call, tokens, cost, and the assistant's response.Additional Comments
Apache 2.0 OSS. Also provides companion App SDKs (@voightxyz/openai, @voightxyz/anthropic, @voightxyz/vercel-ai) for production LLM app monitoring beyond just Claude Code.
Recommendation Checklist