Get This Tool
Caveman
Pricing
- Model
- Free
Summary
Every token your agent burns in a loop you didn't need costs real money — and most terminal coding agents offer no mechanism to stop the bleed. Caveman is a token-reduction stack that sits between your agents and the model, compressing prompts, persisting memory, and gating rollouts behind evals before any savings count.
The vendor claims roughly 65% token reduction across four compression layers — proxy, memory, code, and eval-gated rollout — without altering the bytes the model actually sees. Caveman Code is a terminal agent that plans before it ships, running one autonomous loop across 20+ providers at roughly half the token cost of comparable agents. Cavemem adds a local SQLite store with full-text search and a vector index over MCP, so agents recall prior context instead of re-sending it. The cloud gateway, which would extend these savings across any LLM traffic via a base URL swap, is waitlist-only — it is not available yet. Teams who need the proxy layer today are blocked.
Bottom line: Caveman earns its place in a Claude Code or similar terminal-agent workflow where token costs are already visible on your invoice — but if the cloud gateway is the feature you need, you are on a waitlist with no confirmed ship date.
Community Performance Report Card
No community ratings yet. Be the first to rate this tool!
Community Benchmarks Community
Sign in to submit a benchmarkNo community benchmarks yet. Be the first to share a real-world data point.
Pros
Sign in to edit- Four-layer local compression stack reduces billed tokens without changing what the model sees, so you cut costs without introducing prompt drift or accuracy risk.
- Cavemem persists agent memory in a local SQLite store with vector search over MCP, which means agents stop re-sending full context on every call — the single biggest source of redundant token spend in multi-session workflows.
- Plan-first execution in Caveman Code means the agent ranks moves before shipping, so you avoid the costly re-run loops that happen when an agent discovers mid-task that its approach was wrong.
- MIT-licensed and self-hostable with no API dependency for the local stack, so your token data and agent history stay on your infrastructure rather than routing through a third-party service.
- Cave Architect converts live telemetry into a ranked optimization plan split by dollars-per-day and application change cost, so you prioritize the cheapest wins first instead of guessing where token waste lives.
Cons
Sign in to edit- The cloud gateway — the component that compresses arbitrary LLM traffic via a base URL swap across your whole stack — is waitlist-only and not available. Teams who need organization-wide compression across multiple services or providers, not just terminal agent sessions, have no path forward here and will look at alternatives that ship a working proxy today.
- Caveman Code is a terminal-first agent. Teams building GUI-driven workflows, notebook-based pipelines, or non-terminal agent architectures get no compression benefit from the Code layer — they are limited to the browser extension and whatever the waitlisted proxy eventually delivers.
- The vendor's 65% token reduction claim is stated as live and cited, but it reflects the tool's own metering on its own workloads. Teams with different prompt structures, retrieval patterns, or provider mixes will see different numbers — there is no documented methodology for estimating savings on a novel codebase before committing to integration.
Community Reviews
Sign in to write a reviewNo reviews yet. Be the first to share your experience.
About
- Platforms
- CLI, browser extension, npm packages
- API Available
- No
- Self-Hosted
- Yes
- Last Updated
- 2026-07-11T13:23:25.503Z
Best For
Who it's for
- Developers using Claude Code or similar agents
- Teams seeking lower token costs without accuracy loss
- Agent builders needing memory and compression layers
What it does well
- Reducing token usage in AI coding agents
- Adding persistent memory to agents
- Token-efficient terminal-based coding workflows
- Compressing LLM traffic via proxy gateway
Integrations
Discussion Community
Sign in to commentNo discussion yet. Sign in to start the conversation.
Spotted incorrect or missing data? Join our community of contributors.
Sign Up to ContributeCommunity Notes & Tips Community
Sign in to contributeBe the first to contribute. General notes, observations, gotchas, and tips from people who use this tool day-to-day.
Frequently Asked Questions
- Is Caveman free?
- Caveman has a permanent free tier alongside paid upgrades. You can keep using a baseline version indefinitely without paying.
- Is Caveman open source?
- Yes. Caveman is open source.
- Can I self-host Caveman?
- Yes. Caveman supports self-hosting on your own infrastructure.
- What platforms does Caveman support?
- Caveman is available on: CLI, browser extension, npm packages.
Hours Saved & ROI Stories Community
Sign in to contributeBe the first to contribute. Concrete time/cost savings, with context. e.g. "Cut my code review backlog from 4h to 45m per week."
Curated lists that include this category
AI coding agents running in autonomous loops accumulate token costs fast: every re-sent context window, every redundant plan, every forgotten prior decision compounds. Caveman addresses this with five stacked layers — a compression proxy, persistent agent memory, a terminal coding agent, a telemetry-driven planning tool, and an eval-gated rollout system. Installation is local-first: a curl script for the Claude Code skill, npm packages for Caveman Code and Cavemem, and a browser extension for ChatGPT, Claude, and Gemini contexts. The vendor states the system is byte-safe — the model-visible bytes are never altered, only the billed token count changes.
The differentiating feature is Cavemem, a persistent recall layer served over MCP. Rather than re-injecting full conversation history into each request, agents query a local SQLite database with FTS5 and a vector index to pull only what is relevant. This is the architectural answer to one of the most expensive failure modes in multi-session agent work: agents that have no memory of what they already resolved.
Caveman Code, the terminal agent, adds a planning phase before any code ships, running across 20+ providers in a single autonomous loop. Cave Architect sits above it, turning measured token spend into a ranked list of optimization moves, each tagged with the application change it requires — splitting the decision of what to fix by how much it costs to fix it. Eval-gated rollout then gates any claimed savings behind replay, shadow, and canary stages; the vendor states nothing counts until the eval passes and auto-rollback is armed if quality slips.
The cloud Caveman Proxy — which would compress any LLM traffic via a single base URL swap — is described on the page as ‘coming soon’ with a waitlist. The open-source components (MIT license) install today; the gateway does not. Teams needing organization-wide token metering across multiple services cannot use that layer yet.