README.md
Explore / token-optimizer
alexgreensh
token-optimizer
Context-waste hunter for coding agents: hooks compress reads, bash and search output, checkpoint before compaction, and audit waste in configs, skills, MCP and memory — with a local dashboard.
2,057 161 Python NOASSERTIONupdated 2 days ago
Curator's take
The one token tool that argues with the others in its own README, and mostly wins the argument. Compressors like Headroom and RTK cover command output — roughly 15-25% of your context. Token Optimizer covers eight surfaces (bash, grep, tabular, file-re-read diffs, structure skeletons, archived large results, model verbosity, structural context) and then keeps going: checkpoints before auto-compact so savings survive it, model-routing nudges, loop detection, 30-day trend coaching, and per-component audits of CLAUDE.md, skills and MCP. It is cache-safe, injects nothing into your context, and measures before/after. The costs are real too: it is a large Python/TypeScript surface with 92 env knobs and 15 SQLite tables doing hook surgery on every Read and Bash, its headline dollar figures are counterfactual models against the author's own frozen baseline, and the license is non-standard. Pick RTK if you want one boring binary; pick this if you want the whole waste budget attacked and instrumented.
Mapped by ShipWithAI editors · links verified
Continue your stack
What teams reach for next — and why each earns a place beside token-optimizer. Ranked by curator confidence.
Pairs well with
agentacct
Local-first work receipts for coding agents: reads the session logs Claude Code, Codex, OpenCode and Hermes already write and reports what each task did, what it cost, and whether a check proved it.
Why it fitsTwo readings of the same session logs. Token Optimizer treats them as a budget to shrink — compress, checkpoint, reroute the model. Agentacct treats them as an audit trail: what the task did and whether a check proved it. Run both and you know what the work cost and whether it was worth paying for.
tokensave
Code-intelligence MCP server for coding agents — a pre-indexed semantic graph (libSQL + FTS5) they query instead of grepping: symbols, callers, impact radius in one call. 100% local, 50+ languages.
Why it fitsOpposite ends of file-context cost: Tokensave stops the agent grepping the repo by serving a pre-indexed symbol graph, Token Optimizer shrinks the reads that still happen into diffs and skeletons. Neither duplicates the other.
claude-mem
Cross-harness session memory: hooks capture what the agent does, an LLM compresses it into observations, and the next session gets the relevant ones back via progressive-disclosure MCP tools.
Why it fitsClaude-mem injects recalled observations into every new session; Token Optimizer is the tool that audits what that injection costs, scores MEMORY.md/CLAUDE.md health, and checkpoints the result so it survives auto-compact. One fills the context, the other polices it.
Alternative to See all + compare →
headroom
Context compression layer for agents: squeezes tool outputs, logs, files and RAG chunks 20-95% before the LLM — reversible, local-first; library, proxy, one-command agent wrap, or MCP.
Why it fitsSame goal, different reach. Headroom is a reversible compression layer for tool outputs, logs and RAG chunks, and injects instructions to do it; Token Optimizer compresses more surfaces (search, re-read diffs, structure maps) with zero baseline context overhead, and adds compaction survival and structural audits Headroom does not attempt.
rtk
Rust CLI proxy compressing dev-command output 60-90% before your agent reads it — git, tests, linters, docker, 100+ commands; hooks auto-rewrite bash calls. Single binary, <10ms overhead.
Why it fitsRTK is one Rust binary that compresses shell output pre-shell in <10ms and stops there. Token Optimizer reaches bash plus grep, file re-reads and compaction loss, and reports dollars — at the cost of a much larger installed surface. Pick RTK for minimal, boring and fast.