manifest-build
brunobuddy 作
AI & LLMs概要
Open-source LLM routing and cost tracking plugin.
説明
このスキルでできること
Manifest is an OpenClaw plugin that routes each LLM request to the most cost-effective model using a 23-dimension scoring algorithm with under 2ms added latency. It tracks spend, token usage, and agent activity in a real-time dashboard, and supports hard spending caps with email alerts.
Routing and cost tracking would otherwise require custom code and separate dashboards; Manifest adds both as a no-config plugin with 300+ model coverage.
使いどころ
- Cutting monthly LLM costs by routing simple queries to cheaper models
- Checking which model is consuming the most budget this week
- Setting a hard spending cap before a client project starts
- Asking the agent mid-conversation how many tokens it has used today
- Diagnosing rate limit errors through the message log and error filters
ワークフロー例
User asks: how much have I spent this week and which model cost the most?
- 1
Invokes manifest_costs tool with period set to "week"
- 2
Retrieves cost breakdown by model in USD
- 3
Identifies the highest-cost model from the returned data
- 4
Summarizes total spend and top model for the user
A cost breakdown listing total weekly spend and the most expensive model, pulled from the agent's own usage data
要件
- Manifest API key (mnfst_*) — sign up at app.manifest.build, create an agent, and copy the key
インストール
clawhub install brunobuddy/manifest-buildCLIコマンドを実行するか、Claude Code、Cursor、Codexなどお使いのコーディングエージェントにプロンプトを貼り付けてください。