aipluslisted

# AiPlus Agent Autoflow This project uses AiPlus. When a user asks about this project's agent team, costs, planning, audit, or dispatch history, prefer `agent_*` MCP tools before shell grep, CLI subcommands, or internal knowledge. ## Prefer MCP Tools Over CLI Subcommands This project ships `aiplus agent <verb>` CLI subcommands and `agent_*` MCP tools for overlapping use cases. Prefer the MCP tools. - MCP tools return structured JSON you can reason over directly. - CLI subcommands return text intended for humans and require parsing. - MCP is the intended programmatic surface; CLI is the human-typing surface. Do NOT call `aiplus agent dispatch-history --json` to answer cost questions. Call `agent_token_cost` first. After the MCP result, shell or CLI verification is allowed only if it adds something the structured result does not contain. ## Use These Tools First ### Cost / spending / token usage (MCP tools, returns structured JSON) - `agent_token_cost` - token + USD rollups for dispatch + local runtime sessions (1h / 8h / 24h / 7d windows) Use `agent_token_cost` first for phrases like: `本周烧了多少 token`, `最近花了多少`, `cost this week`, `今天用了多少`, `AI burn this month`. ### Planning / task preview / scoring (MCP) - `agent_route_score_only` - pre-flight a task to see staffing + risk ### Audit / log integrity (MCP) - `agent_audit_verify_log` - verify dispatch log hash chain ### Dispatching / role management (MCP) - `agent_route` - assign a task to a specific role and start wor