See where Claude Code burns tokens — and hit limits less
Observability for Claude Code sessions: per-session and per-file token usage, hot paths that chew through context, and which prompts are quietly the most expensive. Aimed at teams who hit rate limits or want to understand where their Anthropic spend is actually going. PH launch May 2026 with 264 votes — strong traction in the Claude Code dev-tooling category.
| Directr tier | C | S | S | S |
| Price | Free / From $19/mo | From $20/mo or usage-based | From $20/mo | Free (pay API tokens only) |
| Free tier | Personal use | None | 150k tokens/day | Free |
| API access | ✓ | ✓ | ✗ | ✓ |
Anthropic's CLI agent reads your entire codebase, plans multi-file changes, runs tests, and executes complex engineering tasks autonomously from the terminal. Built for engineers who want AI that understands a full project in context rather than autocompleting in isolation — agentic by design and unusually good at not breaking things. Included with Claude Pro ($20/mo) and Max ($100/mo or $200/mo) subscriptions with weekly usage allowances, or pay per token via the Anthropic API.
Build and deploy full-stack web apps from the browser by describing what you want. Bolt writes the code, installs the packages, and runs the server — no local environment required and no setup time. Strong for rapid prototyping and for non-developers who need something working fast rather than a codebase they'll be maintaining long-term.
VS Code extension with Plan/Act modes giving fine-grained control over how autonomously the AI operates. Supports any LLM provider and costs only API tokens with no markup — with file editing, terminal access, browser use, and MCP integration built in. Five million developers and growing fast; the open-source alternative with genuine capability.