Context is editable now — I made a plugin for DeepSeek Harness to manually edit and exclude context

I made Agent Context Editor for DeepSeek Harness , an open-source plugin that lets you manually control what your coding agent actually sees in its context. You can: Search the entire conversation Filter User / AI / Tool messages Select individual messages, ranges, or batches Manually exclude messages from the model's context Manually edit the context content the model sees Restore excluded or edited content later Hide / restore / reset / undo changes Manage AI reasoning and final answers separately Keep the original Session untouched — all context changes are stored separately and are reversible Currently supports Pi TUI and DeepSeek Harness for actual model-context modification, with Pi Desktop also providing the editor UI. One major feature is still unfinished: AI-assisted context cleanup . The idea is to let AI condense selected messages or the whole session, then replace that part of the model context with the approved condensed version. Basically: your context doesn't have to be untouchable anymore. Just edit it. GitHub: github.com/jermaine123123/agent-context-editor Feedback is welcome.