tool_result, and those payloads add up fast. A single Claude Code session can consume tens of thousands of tokens just from tool calls.
Edgee compresses these tool_result messages losslessly before they reach the model. The agent sees identical output. You pay for fewer tokens.
Up to 35% longer sessions for Pro/Max users
Up to 50% lower costs on API billing
Zero config, the CLI handles everything
How It Works
When you launch a coding agent through the Edgee CLI, all traffic is automatically routed through the Edgee gateway. Before each request reaches the LLM provider, Edgee compresses thetool_result content, file contents, grep outputs, bash responses, using lossless techniques that preserve every byte of meaning.
The model receives the same information. You use fewer tokens.
After each session, the CLI prints a link to your session analytics, including tokens used, tokens saved, and cost breakdown.
Supported Tools
| Tool | Command | Status |
|---|---|---|
| Claude Code | edgee launch claude | ✅ Available |
| Codex | edgee launch codex | ✅ Available |
| OpenCode | edgee launch opencode | ✅ Available |
| Cursor | edgee launch cursor | 🔜 Coming soon |
What Gets Compressed
Edgee targetstool_result messages in Claude/Codex/OpenCode… API format:
- File contents: output from Read and file system operations
- Grep and search results: ripgrep, code search, similar tools
- Shell output: stdout/stderr from Bash commands
- API responses: large JSON or text payloads from tool calls
- Database results: rows and records from tool-executed queries
Compression is fully lossless for coding agent traffic. There are no quality thresholds or configuration options, just enable it and go.
Next Steps
CLI Quickstart
Install the CLI and launch your first compressed session in under a minute.
Claude Token Compression
Deep dive into the compression engine powering Claude Code support.
Codex Token Compression
Learn more about how compression is optimized for Codex sessions.