recall
Recall — Restore Developer Memory
When to Use
Call mcp__goldfish__recall to restore context from previous sessions. Recall runs automatically at session start, but users can also invoke /recall for targeted queries (search, cross-project, specific time ranges).
mcp__goldfish__recall({})
Default parameters (last 5 checkpoints, no date window) cover most cases.
Common Scenarios
- New session, need prior context —
recall()with defaults - After context compaction — recall to restore lost state
- Searching for past work —
recall({ search: "auth refactor", full: true }) - Cross-project standup —
recall({ workspace: "all", days: 1 }) - Just need the plan —
recall({ limit: 0 })
Parameter Examples
Standard recall (most common)
mcp__goldfish__recall({})
Need more history
mcp__goldfish__recall({ days: 7, limit: 20 })
Looking for specific work
mcp__goldfish__recall({ search: "auth refactor", full: true })
Recent activity only
mcp__goldfish__recall({ since: "2h" })
Interpreting Results
Recall returns up to three sections:
1. Active Plan (top of response)
The current strategic plan for this workspace. If present, work should align with it.
2. Checkpoints (chronological array)
Each checkpoint contains:
timestamp— when it happened (UTC)description— what was done, why, and howtags— categorization labelsgit.branch,git.commit— git state at checkpoint time (only withfull: true)git.files— changed files (only withfull: true)
3. Workspace Summaries (cross-project only)
When using workspace: "all", you get per-project summaries with checkpoint counts.
Processing Large Result Sets
When you get 10+ checkpoints back, distill them:
- Group by date — what happened each day
- Identify themes — feature work, bug fixes, refactoring, planning
- Highlight blockers — anything marked stuck, blocked, or failed
- Surface decisions — architectural choices, tradeoffs made
- Find the thread — what was the user working toward?
Present a concise summary: "Based on your recent work, you were [doing X] on [project area]. Last session you [accomplished Y] and the next step appears to be [Z]."
After Recall
Once you have context, act on it:
- Recall (restore memory)
- Understand (process what you get back)
- Continue (pick up where the last session left off)
Trust recalled context — don't re-verify information from checkpoints.
More from anortham/goldfish
plan
Create and manage persistent plans in Goldfish memory — use when starting multi-session work, making architectural decisions, or when the user discusses project direction, roadmaps, or design decisions that need to persist across sessions
13plan-status
Assess progress against the active Goldfish plan using checkpoints and plan data — use when the user asks about project progress, how things are going, what's been accomplished, or wants a status check against their plan
12checkpoint
Save developer context to Goldfish memory — checkpoint at meaningful milestones, not after every action
12standup
Generate a standup report from Goldfish memory across all projects — use when the user asks for a standup, daily update, progress summary, what they've been working on, or needs a report for a team sync
11session-memory
Automatically restore session context from persistent memory at the start of every coding session. Use this skill IMMEDIATELY when starting work to recall previous checkpoints, active plans, and git context. MANDATORY for context restoration.
2progress-tracking
Proactively checkpoint work progress at key moments without asking permission. Use when completing tasks, fixing bugs, making discoveries, or reaching milestones. Builds persistent memory across sessions.
2