Token Usage, Token Cost and Tool Prompting

[0.21.0] - 2026-02-28

Feature/token-usage, token-cost (via ratecard)

• Token usage visibility per LLM call and response
• Token stats provider per LLM call, responding with input and output tokens for the turn call
• Agent has input_tokens: int = 0 and output_tokens: int = 0 these provide an accumulative count
• Implemented ratecard for each provide and model so the actual token usage cost can be returned to agent

[0.20.0] - 2026-02-25

Feature/toolprompt

• Problem: We want human-in-the-loop for tool invocations
• agentify run agent.yaml --toolprompt will force the agent to prompt before making any tool calls
• This is basic control for now, can be evolved to trust levels of agents and tools.