| name | load-context |
| description | Load all LTM memories for the current session. Primary method when SessionStart hook is disabled (Windows Terminal bug workaround). |
Load Context
Manually loads all LTM memories into the current session context.
When to Use
- Windows Terminal workaround: When the SessionStart "startup" hook is disabled due to Claude Code issue #23083 (stdin bug on Windows Terminal)
- Context refresh: When you want to reload memories mid-session
- Alias:
/refresh-memories does the same thing
Usage
Use the MCP tool (works in all projects):
mcp__anima__memory(action="refresh")
Or via CLI (Anima project only):
uv run anima load-context
What It Does
- Resolves current agent (Anima) and project
- Loads ALL memories at once (no deferred loading -
--all flag is default)
- Checks for recent dreams and cognitive dissonances
- Checks curiosity queue for research prompts
- Outputs formatted memory block for context injection
Note: When called from CLI, --all is automatic. No need to run /load-deferred separately.
Subconscious memories are now indexed automatically at session end.
Use /recall --subconscious to search past dialogues.
Configuration
To disable automatic memory injection (workaround for Windows Terminal issue #23083):
- Remove the "startup" matcher from SessionStart hooks in
.claude/settings.local.json
- Start Claude Code normally (no freeze)
- Run
/load-context to load memories manually
The other matchers (resume, compact, clear) can remain - they fire after the session is interactive.
Output
Returns the full session start context including:
- Memory DSL block with all loaded memories
- Dream prompts (if recent dreams)
- Dissonance alerts (if contradictions need resolution)
- Curiosity prompts (if research is due)
Difference from /load-deferred
/load-context: Loads ALL memories (full session start)/load-deferred: Loads only the memories that were deferred due to 25KB hook output limit
Use /load-context when auto_inject is disabled.
Use /load-deferred after greeting when auto_inject is enabled.
