Unverified50% confidenceFactExact time
Coding agents typically work in a multi-turn loop, resending the full conversation history (including the skill file in the system prompt) on every turn, so every KB of the skill file is billed repeatedly on every request, and the cost scales linearly with the number of loop turns.
1
Sources
50%
Confidence
Long-term
Relevance
10/18/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedA coding agent resends the entire conversation history every turn, which is the fundamental reason input tokens accumulate quickly.76% similarUnverifiedCoding agents stuff entire code files, function dependencies, and conversation history into the context window to send to the model, causing a single refactoring operation to potentially trigger dozens of API calls, with cumulative token consumption reaching several to tens of times that of manual coding.74% similarUnverifiedAgent类应用因每轮对话携带几乎相同的系统提示词和追加历史记录，前缀重叠率高，天然适合缓存优化以降低成本72% similarUnverifiedIn complex Agent tasks, multi-turn conversations and tool calls rapidly accumulate context length, directly affecting API costs and response speed; token costs can be managed through strategies such as conversation compression and summary caching.70% similarUnverifiedThe observation returned by any tool call permanently resides in the conversation context as tokens, and is re-read and billed in every subsequent round, so the more verbose the tool output, the higher its marginal cost.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1062640API
curl https://kongchang.com/api/v1/knowledge/claims/1062640MCP
get_claim(id=1062640)