What problem does this solve?
TokenTracker currently works well for tracking token usage from CLI tools and supported providers, but it lacks:
- visibility into agent-based workflows (e.g., Craft Agent), where token usage happens across multiple steps and tool calls
- support for newer LLM providers like DeepSeek, which are increasingly used for cost-efficient inference
This makes it harder to use TokenTracker in more advanced or production AI setups.
Proposed solution
- Add basic integration support for agent frameworks like Craft Agent:
- track token usage per task / step
- allow passing simple metadata (e.g., task_id, agent_name)
- Add DeepSeek API support:
- support deepseek-v4
- track token usage and cost via OpenAI-compatible interface
A lightweight implementation (e.g., wrapper or adapter) would already be very helpful.
⸻
Thanks for the great project!
Alternatives considered
No response
Category
New AI tool integration
Anything else?
No response
What problem does this solve?
TokenTracker currently works well for tracking token usage from CLI tools and supported providers, but it lacks:
This makes it harder to use TokenTracker in more advanced or production AI setups.
Proposed solution
A lightweight implementation (e.g., wrapper or adapter) would already be very helpful.
⸻
Thanks for the great project!
Alternatives considered
No response
Category
New AI tool integration
Anything else?
No response