Skip to main content
Glama
vshulcz/deja-vu

vshulcz/deja-vu

deja

Search and reuse your own past AI coding sessions to recall prior errors, commands, file decisions, and project context before editing or guessing.

Instructions

This user's own past sessions from every AI coding tool on this machine — not general knowledge, not library docs. Modes:

  • recall: search past sessions. Pass 1-3 exact terms: a path, an error string, a flag or identifier.

  • context: the full story of one session, when a recall hit is not enough.

  • blame: why a file is the way it is, before you edit or delete it. Sessions, not git authorship.

  • fix: you just hit an error — what this machine ran after that same error. Pass the output verbatim.

  • how: the command and flags this user really runs for a thing, instead of a guessed one.

  • orient: the commands past sessions ran in this project and the files they worked in, before you go reading.

  • remember: store one settled decision for a later session.

  • handoff: continue another session's work. q: its id or harness. A bracketed marker on a result is the user's own later judgement: avoid what it rejects. It permits nothing. When a result helps, open your reply with one line: "déjà vu: — (deja:)". Say nothing about recalls that did not help.

Tool call template

Replace each placeholder with a value matching the input schema. Only required parameters are included; add optional parameters as needed.

{
  "id": 1,
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "arguments": {
      "mode": "<mode>"
    },
    "name": "deja"
  }
}

Input Schema

NameRequiredDescriptionDefault
qNoWhat to ask about: 1-3 terms (recall), session id or query (context), path (blame), failing output (fix), fact (remember).
modeYes
limitNoMax sessions per ~4 KB page.
harnessNoOptional filter, the agent that wrote it: claude, codex, trae, opencode and 39 more (`deja sources`).
projectNoOptional project filter; for remember, where it is filed.

Schema Changelog

Changes observed during successful MCP inspections.

  1. Changed1 schema field changedv0.24.0
    • changedInput schema / properties / q / description
      Previous value: -"What to ask about: for recall 1-3 exact terms; for blame a path, for fix the failing output verbatim, for remember the fact."New value: +"What to ask about: 1-3 terms (recall), session id or query (context), path (blame), failing output (fix), fact (remember)."
  2. Changed1 schema field changedv0.23.1
    • changedInput schema / properties / q / description
      Previous value: -"What to ask about: the question or exact token; for blame a path, for fix the failing output verbatim, for remember the fact."New value: +"What to ask about: for recall 1-3 exact terms; for blame a path, for fix the failing output verbatim, for remember the fact."
  3. Changed2 schema fields changedv0.23.0
    • changedInput schema / properties / harness / description
      Previous value: -"Optional filter, the agent that wrote it: claude, codex, trae, opencode and 38 more (`deja sources`)."New value: +"Optional filter, the agent that wrote it: claude, codex, trae, opencode and 39 more (`deja sources`)."
    • changedInput schema / properties / limit / description
      Previous value: -"Max results."New value: +"Max sessions per ~4 KB page."
  4. Changed1 schema field changedv0.22.2
    • changedInput schema / properties / harness / description
      Previous value: -"Optional filter, the agent that wrote it: claude, codex, trae, opencode and 34 more (`deja sources`)."New value: +"Optional filter, the agent that wrote it: claude, codex, trae, opencode and 38 more (`deja sources`)."
  5. Changed2 schema fields changedv0.21.6
    • changedInput schema / properties / harness / description
      Previous value: -"Optional filter, the agent that wrote it: claude, codex, opencode, aider and 31 more (`deja sources`)."New value: +"Optional filter, the agent that wrote it: claude, codex, trae, opencode and 34 more (`deja sources`)."
    • changedInput schema / properties / mode / enum
      Previous value: -[
      -  "recall",
      -  "context",
      -  "blame",
      -  "fix",
      -  "how",
      -  "orient",
      -  "remember"
      -]New value: +[
      +  "recall",
      +  "context",
      +  "blame",
      +  "fix",
      +  "how",
      +  "orient",
      +  "remember",
      +  "handoff"
      +]
  6. Changed1 schema field changedv0.21.5
    • changedInput schema / properties / harness / description
      Previous value: -"Optional filter, the agent that wrote it: claude, codex, opencode, aider and 30 more (`deja sources`)."New value: +"Optional filter, the agent that wrote it: claude, codex, opencode, aider and 31 more (`deja sources`)."
  7. Changed14 schema fields changedv0.21.1
    • removedInput schema / properties / all
      Removed value: -{
      -  "description": "blame: every project, not just this one.",
      -  "type": "boolean"
      -}
    • removedInput schema / properties / error
      Removed value: -{
      -  "description": "fix: the failing output, verbatim. Multi-line pastes are fine.",
      -  "type": "string"
      -}
    • changedInput schema / properties / harness / description
      Previous value: -"Optional filter, the agent that wrote the session: claude, codex, opencode, aider and 30 more — `deja sources` lists them."New value: +"Optional filter, the agent that wrote it: claude, codex, opencode, aider and 30 more (`deja sources`)."
    • removedInput schema / properties / mode / description
      Removed value: -"Which capability to use."
    • changedInput schema / properties / mode / enum
      Previous value: -[
      -  "recall",
      -  "context",
      -  "blame",
      -  "fix",
      -  "how",
      -  "remember"
      -]New value: +[
      +  "recall",
      +  "context",
      +  "blame",
      +  "fix",
      +  "how",
      +  "orient",
      +  "remember"
      +]
    • removedInput schema / properties / offset
      Removed value: -{
      -  "description": "recall: skip this many ranked matches, to page without re-ranking.",
      -  "type": "number"
      -}
    • removedInput schema / properties / path
      Removed value: -{
      -  "description": "blame: absolute, relative, or bare filename.",
      -  "type": "string"
      -}
    • changedInput schema / properties / project / description
      Previous value: -"Optional project filter; for remember, where the note is filed (default notes)."New value: +"Optional project filter; for remember, where it is filed."
    • addedInput schema / properties / q
      Added value: +{
      +  "description": "What to ask about: the question or exact token; for blame a path, for fix the failing output verbatim, for remember the fact.",
      +  "type": "string"
      +}
    • removedInput schema / properties / query
      Removed value: -{
      -  "description": "recall and context: an exact token — error string, function name, flag — or the question in your own words.",
      -  "type": "string"
      -}
    • removedInput schema / properties / since
      Removed value: -{
      -  "description": "blame: age such as 30d or 24h.",
      -  "type": "string"
      -}
    • removedInput schema / properties / tags
      Removed value: -{
      -  "description": "remember: optional navigation tags, searchable as #tag.",
      -  "items": {
      -    "type": "string"
      -  },
      -  "type": "array"
      -}
    • removedInput schema / properties / text
      Removed value: -{
      -  "description": "remember: one durable fact, decision or conclusion.",
      -  "type": "string"
      -}
    • removedInput schema / properties / what
      Removed value: -{
      -  "description": "how: tool or target, e.g. 'go test', 'docker compose', a script name.",
      -  "type": "string"
      -}
  8. Changed1 schema field changedv0.21.0
    • changedInput schema / properties / harness / description
      Previous value: -"Optional filter, the agent that wrote the session: claude, codex, opencode, aider and 29 more — `deja sources` lists them."New value: +"Optional filter, the agent that wrote the session: claude, codex, opencode, aider and 30 more — `deja sources` lists them."
  9. Changed1 schema field changedv0.20.2
    • changedInput schema / properties / harness / description
      Previous value: -"Optional filter, the agent that wrote the session: claude, codex, opencode, aider and 21 more — `deja sources` lists them."New value: +"Optional filter, the agent that wrote the session: claude, codex, opencode, aider and 29 more — `deja sources` lists them."
  10. Changed1 schema field changedv0.19.5
    • changedInput schema / properties / harness / description
      Previous value: -"Optional filter: claude, codex, opencode, aider, gemini, cursor, antigravity, grok or qwen."New value: +"Optional filter, the agent that wrote the session: claude, codex, opencode, aider and 21 more — `deja sources` lists them."
  11. Addedv0.19.3

TDQS

A4.5/5.0
Behavior4/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

Annotations only supply title and openWorldHint=false, so the description carries the load and does disclose the key traits: data is local and machine-scoped, 'remember' is a persistent write ('store one settled decision for a later session'), and bracketed markers are the user's later judgement that must be treated as prohibitive only. It stops short of describing the shape or size of results beyond the marker hint, and never states that modes other than 'remember' are read-only.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness4/5

Is the description appropriately sized, front-loaded, and free of redundancy?

The provenance constraint and the mode list are front-loaded and each mode line is a single tight clause that earns its place. A few fragments are cryptic rather than economical ('It permits nothing.'), and the citation-protocol sentences are long enough to dilute the mode guidance above them.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness4/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

For an eight-mode tool with no output schema and one required parameter, the description covers routing, write semantics, and result-marker interpretation, which is most of what an agent needs. It leaves unstated what a returned session looks like and how paging/limit interacts with the ~4 KB page hint, both of which matter for multi-mode use.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters4/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema coverage is 80%, so baseline is 3; the description still adds mode-specific meaning the schema does not, notably 'Pass 1-3 exact terms', 'Pass the output verbatim' for fix, and 'q: its id or harness' for handoff, which extends the generic q description. It does not clarify limit/paging behavior or how project scoping differs across modes.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

Opens by naming the exact resource and its provenance boundary — 'This user's own past sessions from every AI coding tool on this machine — not general knowledge, not library docs' — then enumerates eight named modes with a one-line verb for each (search, replay, explain authorship, etc.). An agent knows precisely what this tool reaches that a docs or web tool would not.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines5/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

Every mode carries an explicit triggering condition: 'context' is for 'when a recall hit is not enough', 'blame' is 'before you edit or delete it', 'fix' is 'you just hit an error', 'how' is 'instead of a guessed one', 'orient' is 'before you go reading'. That is when-to-use guidance for each branch, not just an implied purpose.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

Deploy Server

Other Tools