Testing 3 more Claude Code cost-cutting plugins: LSP saves 13%
Marmelab · reddit · 2026-09-16
Following an earlier round of testing, a Marmelab engineer tested 3 more Claude Code cost-cutting plugins on a multi-agent internal CRM build (orchestrator + dev + reviewer agents), where the token hog turned out to be cache reads, not generated code.
- LSP (clear winner): native LSP support gives agents precise code navigation instead of blind grep. Cut costs 13%, tokens 12%, API calls 24%; grep/rg invocations dropped 44% (340 → 189), replaced by 55 precise LSP lookups. Caveat: LSP doesn't work in subagents, so it must stay at the top level.
- Graphify (net negative): builds a project dependency graph via static analysis + an LLM pass; the extra LLM overhead made it slightly more expensive than baseline.
- RTK (wrong fit): wraps the CLI to compress shell output; helpful only if verbose CLI output is a major token source — it wasn't here.
Bottom line: the best combo is Ponytail (agents write less code, less reviewer ping-pong) + LSP. The author declares no affiliation with any plugin.
More from coding & agent
- Microsoft's MemSteer turns unlabeled user histories into agent memory supervision — MicrosoftResearch · 2026-11-02
- She fired herself from her own Instagram: AI agent runs the account, views up 82.3% — heyshrutimishra · 2026-11-02
- Sourcegraph: AI writes code faster than teams can review it, driving more outages — AI Engineer · 2026-11-02
- Tsinghua & Microsoft's HarnessMM cuts agent token use by up to 87.93% while raising success rate — rohanpaul_ai · 2026-11-02
- GitHub launches Agent HQ: Anthropic, OpenAI, Google coding agents in Copilot — emmanuelvivier · 2026-11-02
- DeepSeek open-sources Super-Agent, a coding agent that reads team chats to minimize human input — emmanuelvivier · 2026-11-02