You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Add agent memory — persistent per-file context for AI sessions
Agents can now save and read a memory note tied to the current log
file so analysis can resume where it left off across sessions.
Storage: <logdir>/.logan/<filename>.agent-memory.json sidecar file.
Auto-inject on connect: when a named agent connects via /api/events
the SSE connected event includes the current memory payload so the
agent gets prior context without explicitly calling logan_memory_read.
MCP tools:
logan_memory_read — read the saved memory for the current file
logan_memory_write — write/update the memory note
API endpoints:
GET /api/agent-memory — read
POST /api/agent-memory — write (content + optional agentName)
POST /api/agent-memory-clear — delete
IPC: agent-memory-get/save/clear + agent-memory-changed push event.
UI: a compact memory bar in the Chat panel shows the agent name,
time ago, and a one-line preview of the last saved note. Disappears
when no memory exists for the current file; ✕ button to clear.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
0 commit comments