Index and recall coding decisions across all AI agents

curl -fsSL https://raw.githubusercontent.com/vshulcz/deja-vu/main/install.sh | sh
deja install --auto

Highlights

• Starts full. Months of history from before you installed it are searchable on day one: deja "connection pool exhausted" over gigabytes.
• One memory, every agent. A fix found in Codex comes back in Claude Code, Cursor or opencode; all thirty-five agents read the same index.
• Nobody has to ask. Recall arrives at session start, before a file is edited or a command runs, and after a command fails.
• Survives compaction. Over 43 measured compactions the summary kept 77% of the decisions and 0.2% of the commands; deja hands back the rest (how).
• Indexes the work, not just the talk. The files each turn opened, the commands with their exit status, the exact spans an edit replaced.
• Knows what held. deja promote <id> --state rejected marks a reverted decision, and every later hit says it was tried and why it was dropped; --state accepted takes the mark back.
• Says when the ground moved. A hit reports 4 files this session touched have changed since, and stays quiet when it cannot tell.
• Local and private. No model, no embeddings, no server. Keys and tokens are stripped as the index is built (privacy).
• Moves with you. deja sync ssh laptop between machines, no cloud in between; deja handoff --to codex to continue in another agent.
• One Go binary. macOS, Linux and Windows, through Homebrew, Scoop, winget, npm or go install.

Your own work, wrapped

deja stats --card draws it in the terminal; give it a filename and it writes an SVG for a profile README. To post it anywhere else, turn it into a PNG - that page converts it in your own browser.

The full feature reference lives in the docs.

Install

The two commands at the top are the whole install on macOS and Linux:

curl -fsSL https://raw.githubusercontent.com/vshulcz/deja-vu/main/install.sh | sh
deja install --auto

Ten seconds to install, about ten to index, and it is useful. The second command wires MCP recall into every agent it finds, turns on session-start recall where the agent supports it, and builds the first index so the next session does not pay for it.

Start a new agent session and ask it something you worked on months ago:

have we dealt with jwt refresh rotation before? check your memory

It does not have to be asked, either - with auto-recall the agent already knows what you solved in that project when the session opens.

brew install deja-vu, go install github.com/vshulcz/deja-vu/cmd/deja@latest, or npx @vshulcz/deja-vu "query" to try it without installing anything. Desktop apps that take MCP servers as bundles can open the .mcpb from the latest release; it carries the binary.

Claude Code, Cursor, Qwen and OpenClaw can take the same plugin bundle from their own marketplaces instead (Codex has a bundle of its own, in the table under Harnesses with a package of their own). Copilot CLI installs it too but takes only the skill, so use deja install copilot-auto there:

claude plugin marketplace add vshulcz/deja-vu && claude plugin install deja-vu@deja-vu

On Windows the install script exits with unsupported OS - it is a shell script. Use Scoop, from the main bucket every Scoop install already has, or winget, which ships with Windows:

scoop install deja-vu
winget install vshulcz.deja-vu

Or take deja-vu_<version>_windows_amd64.zip from the latest release and put deja.exe on your PATH, e.g. in %USERPROFILE%\.local\bin.

The binary alone is a complete install for searching: index, search, show, ctx, blame, --json and redaction need nothing else. deja install is what wires MCP into your agents and turns on session-start recall - worth having, and optional. On a binary-only setup deja doctor reports every MCP target as not-wired, which is that setup working as intended. deja warmup also leaves a skill at ~/.agents/skills/deja-search/SKILL.md that teaches an agent the CLI contract - deja search --json, ctx, blame, how to read tier and total - so it knows history is searchable without MCP. The copy in the repo is skills/deja-search/SKILL.md.

deja install --all is --auto without the session-start recall: agents answer from memory when they decide to call it, rather than starting each session with it. The agent setup guide covers what each harness supports, aider's read-only context file, and the Windows cmd /c deja mcp wrapper.

Install also writes user-level guidance for the harnesses it detects: Claude Code, Codex, opencode, Gemini CLI, Antigravity, Qwen, Kimi Code, pi, Senpi, Copilot, VS Code Copilot Chat, Cursor, Goose, OpenClaw, Hermes, Roo Code, omp, Amp, prime-agent, DeepSeek Harness, Continue, Crush and Zed each get it in their own guidance file (or under the configured XDG_CONFIG_HOME). Re-run rewrites deja's skill or marked block without changing surrounding user content. Use deja install --all --no-guidance to opt out; Grok Build gets the shared skill in ~/.agents/skills, which is what it reads; the ~/.grok/GROK.md written beside it is for the unrelated community CLI that shares that directory. Cursor has no user-level instructions file, so it gets the shared skill in ~/.agents/skills - one of the four places Cursor reads skills from - read only when something looks relevant rather than every session. Kilo Code, gajae-code, Command Code, Cherry Studio and Reasonix get the skill, and Kiro a steering file, from their own install target.

CLI

$ deja "jwt refresh token"
[claude] api        · Jul 8 · 8f31c0a9 — 2 matches
  login started failing after refresh token rotation; jwt kid mismatch in tests
  fixed by reloading jwks cache after rotateKey and adding a clock-skew test
[codex]  web        · Jul 1 · b77d91e2 — 1 match
  refresh token cookie needed SameSite=Lax in local callback flow

Ask your history

Use what it finds

Move it and check it

Full reference: commands and JSON output.

MCP tools

The server exposes one tool, deja, with a mode; deja install wires it in. One tool costs 477 tokens of definitions a turn, against 8,283 for the largest of the seven servers measured in day zero. The six older tool names (recall, recall_context, blame, fix, how, remember) still answer.

q carries whatever the mode asks about. The per-mode names below are still accepted; they are no longer declared, because the schema is read every turn whether or not the tool is called.

Supported harnesses

aider · Amp · Antigravity · Claude Code · Cline · Codex CLI · Copilot CLI · VS Code Copilot Chat · Cursor · DeepSeek Harness · Gemini CLI · Goose · Grok Build · Hermes · Kimi Code · omp (Oh My Pi) · OpenClaw · opencode · Continue · Crush · pi · prime-agent (PrimeIntellect) · Qwen Code · Cherry Studio · Senpi · gajae-code · Kimchi Coding · Command Code · ZCode · Kiro · Kilo Code · Roo Code · Zed · CodeWhale · Reasonix.

✅ works · - possible, not built yet · ✕ the harness has no such mechanism · ⚠ waiting on the harness itself · ? not investigated

Custom store locations go through DEJA_*_ROOT variables, and each agent's own relocation variable is honored too. The session format registry documents the observed paths, record schemas and role mapping per harness, with synthetic fixtures keeping those descriptions checked against the parsers.

Harnesses with a package of their own

deja install --auto wires every one of these like every other harness, and that stays the shortest path. They also have a package in their own ecosystem, for people who install extensions there rather than from a CLI:

Either path works alone, and both together double nothing: each package reads what deja install already wrote and uses the deja you already have.

The same search is also a skill, for any agent that loads a SKILL.md:

npx skills add https://github.com/vshulcz/deja-vu --skill deja-search   # skills CLI: Claude Code, Cursor, Goose, Copilot…
openclaw skills install @vshulcz/deja-search                            # ClawHub
hermes skills install vshulcz/deja-vu/skills/deja-search                # Hermes

The skill drives the deja binary from the install step above; it does not bundle one.

Semantic recall (optional)

Point deja embed at a local Ollama, LM Studio or OpenAI-compatible endpoint with DEJA_EMBED_URL and rephrased queries still hit. Without a reachable runtime, lexical search and MCP recall continue unchanged. OpenAI Platform works with its standard key:

export OPENAI_API_KEY='sk-...'
export DEJA_EMBED_URL='https://api.openai.com/v1/embeddings'
export DEJA_EMBED_MODEL='text-embedding-3-small'
deja embed

With no DEJA_EMBED_URL set, deja probes localhost:11434 and localhost:1234, so a machine already running Ollama or LM Studio is picked up without being asked. DEJA_EMBED_OFF=1, or DEJA_EMBED_URL=off, turns that probe off - any other configured DEJA_EMBED_URL still wins.

For another authenticated OpenAI-compatible endpoint, set DEJA_EMBED_KEY explicitly:

export DEJA_EMBED_URL='https://example.com/v1/embeddings'
export DEJA_EMBED_MODEL='embedding-model'
export DEJA_EMBED_KEY='...'
deja embed

DEJA_EMBED_KEY takes precedence. OPENAI_API_KEY is used automatically only for an HTTPS api.openai.com URL; it is never implicitly sent to local or third-party endpoints.

The sidecar sits beside the index as .vectors.bin, not inside index.db. Float32 vectors cost roughly 4 MB per 1k messages for a 1,024 dimension model. A remote endpoint receives the redacted indexed text, truncated to about 2k characters, but never raw source files. With Ollama or LM Studio, embedding stays local and needs no key.

Proof

Millisecond lookups, and on LongMemEval-S 88.1% hit@1 (470-question cleaned set) and 87.4% hit@1 on all 500 questions; on LoCoMo retrieval, 70.5% hit@1. On a task this machine had already solved, 58% fewer tokens: eleven runs an arm, 53,558 against 126,222 with nothing wired, and 52,815 against 103,443 on a later build with the arms alternated. Both retrieval harnesses ship in this repo and run on the public datasets in minutes: benchmarks · what one task costs.

The rest is measured by deja bench:

deja bench recall     # ranking floor: 100 queries, half Russian, CI fails if recall drops
deja bench context    # 30 seeded task chains plus five negative controls
deja bench block      # does the answer survive into what deja hands over
deja bench prompt     # what the per-prompt hook fires on, and what it fires on wrongly
deja bench ingest     # what an update costs: unchanged, a turn, a new transcript, a rename, a rewrite
deja bench read       # what it costs to read a database-backed store, and what one long value does to it

bench block asks the question the others cannot: with the right session in hand, does the block carry what that session settled. Eight sessions discuss each subject and one of them settles it, in the middle of its own transcript rather than at the end - so the baseline arm, the newest turns of the top hit, scores zero and an arm above zero had to choose.

The context experiment compares deja-recall against full-history, naive grep and cold context. With the default seed:

Same fact coverage as grepping the raw logs for about 250x fewer tokens, and about 70x fewer than replaying the matched sessions in full, while injecting nothing on the chains where no prior fact is relevant. The corpus generator and the relevance labels are ordinary reviewed Go. Audit what "relevant" means before trusting any figure, ours included.

Measured on a real store of 2,419 sessions and 179k messages, 1.9 GB of transcripts:

The same store has since grown to 2,754 sessions, 358k messages and 5.7 GB. A cold full build over it takes 71 s and writes a 232 MB index - 4% of the corpus, because the share falls as transcripts repeat themselves - and the end-to-end median is unchanged at 0.25 s.

The index is incremental. When a session file grows, only that file is re-read.

How it works

Local inverted index in ~/.cache/deja: parse the JSONL and SQLite stores, redact credentials, write records.bin plus token buckets, and track per-file state in manifest.gob so repeat runs only ingest what changed. The MCP server, stats, share and sync all read that one index. Details in docs/ARCHITECTURE.md.

FAQ

No, unless you ask it to. See the data flows.

They stay in the original harness files, which are your agent's data; deja secrets names the sessions that carry them so you can rotate and delete, and --scrub rewrites the transcripts it can reach. Known shapes - AWS keys, api_key=/token= assignments, bearer tokens and bare JWTs, PEM blocks, provider tokens, high-entropy values - are stripped as the index is built, so they do not reach digests, shares or sync exports. Pattern matching is not secret detection: a shape it does not know can pass through. See the security model.

A recall is a lexical lookup against a local index: 0.7-0.8 ms median, and nothing waits on a model. A hook adds the process start and a freshness check over your stores on top of that - tens of milliseconds on a store of a few gigabytes.

No. The agent calls recall itself, and with auto-recall it already knows the project's prior decisions when the session opens.

engram is the strongest of the record-forward tools and worth your time if that model fits you; it still starts empty and knows only what an agent chose to save. The full comparison covers 15 of them.

Under ~/.claude/projects, one JSONL file per session; Codex keeps ~/.codex/sessions, Cursor a SQLite state.vscdb. deja search reads them all in place, deja last lists the recent sessions of every agent, and deja view opens the whole history as one local page. Paths for each agent: where sessions are stored.

Claude Code deletes transcripts older than 30 days (cleanupPeriodDays in ~/.claude/settings.json), and claude --resume lists only what is left. A session deja indexed before the cleanup stays searchable after the file is gone. Details: session files on disk.

Builds exist and CI runs the suite there. macOS and Linux are the battle-tested paths. Field reports welcome in #9.

deja uninstall --all
rm -rf ~/.cache/deja

Guides

Written for the situation rather than the feature:

• Does your agent remember previous conversations? - what each agent keeps between sessions, and what it drops
• Session files on disk - how big ~/.claude/projects gets, and what deleting it costs
• The agent lost the context you had - after a crash, a clear, or a session that came back empty
• The context window is full - what compaction keeps, measured, and what to do instead
• Resuming yesterday's session - find it across every agent, reopen it in the one that owns it
• The agent repeats a mistake you already fixed
• Finding the session where you solved it
• Why agents forget between sessions · Where each agent keeps its history
• What compaction drops · Switching agents · Auditing an agent · Exporting a conversation · Across machines · What memory costs

Per harness: opencode · DeepSeek Harness · Kimi Code · Zed · Grok Build · Gemini CLI · Qwen Code · OpenClaw · Goose · Cline · pi and omp · Hermes

Try it on your own history

curl -fsSL https://raw.githubusercontent.com/vshulcz/deja-vu/main/install.sh | sh
deja install --auto

Ten seconds to install, about ten to index. The next session your agent opens, it already knows what you solved in that project - including everything from before you installed this.