About Openwolf
OpenWolf is a local project memory and context tool for coding agents. It keeps task context, a project map and known fixes in a folder beside the code, so that Claude Code, Codex and OpenCode sessions can continue earlier work instead of starting from scratch.
It installs as an npm package and is set up per project with an init command. Checkpoints record the task objective, open problems and next action, and supported Claude and Codex hooks can restore that context at the next session boundary. A project map holds file descriptions, symbols and line ranges, so an agent can search for the relevant code and read only part of a file. Repeated reads of unchanged files are flagged, and on supported Claude hooks large command output can be shortened with the full text kept in a local cache. Handover packets can be exported from a Claude session and imported into a Codex session, or the reverse, and the vendor says they carry saved messages and results, not private reasoning. A dashboard shows memory, recorded token usage by agent and model, and activity history. The vendor states that it does not run a second model to maintain memory or write summaries.
It is aimed at developers who run long or repeated sessions with coding agents and want saved context they can review as files.
The vendor publishes no pricing page. The tool is installed from npm.