goosedump
goosedump browses, searches, converts, and compacts saved coding-agent
sessions. It also provides project-scoped durable memory for Claude Code,
Codex, Crush, Gemini CLI, Goose, OpenCode, and Pi.
It builds on ideas from lllyasviel/VCC.
Install
Prebuilt binaries support Linux x64 and arm64, macOS arm64, and Windows x64:
To build from source:
Requires Node.js 18 or later. Source builds require Rust 1.95 and a C compiler.
Quick start
Commands that address a saved session require a provider:session-id target.
See goosedump --help and the manual page for options.
Compaction and memory
Compaction creates a summary and does not change durable memory. Summaries have
an estimated 4,096-token limit by default; pass --summary-max-tokens 0 to
disable it.
memory learn stores evidence first, then derives cited atomic claims from new
session entries. Recall searches the current project by default, combining
lexical, entity, relation, and semantic matches. Use --history for superseded
claims or --all-projects to search every project.
Recall JSON includes match_reasons and cited evidence records with stable
citation_id values and snippets.
The text and embedding models run locally and download on first use. Evidence, claims, entities, relations, and vectors are stored together in SQLite.
Pi extension
The bundled pi-goosedump extension adds session search, memory, session management, and compaction to Pi:
Development
License
goosedump: LGPL-2.1-or-laterpi-goosedump: Apache-2.0- GPT-OSS-20B: Apache-2.0
- bge-small-en-v1.5: MIT
Models are downloaded at run time and are not distributed with goosedump.