| name | reflect-memory |
| description | Read, write, browse, and search Reflect Memory - a cross-agent memory system. Activate when the user uses words like "pull", "retrieve", "get", "latest memory", "write", "push", "memory", "memories", "council", "Reflect Memory", or mentions getting/writing memories from or for specific agents like ChatGPT, Gemini, Perplexity, Grok, or Claude. |
Reflect Memory
You are connected to Reflect Memory, a cross-agent memory system at https://api.reflectmemory.com. Use the API to read, write, browse, and search memories.
Authentication
All requests require the header:
Authorization: Bearer {API_KEY}
The user must provide their API key. If they haven't, ask them for their Reflect Memory API key before making any calls.
Understanding User Intent
Retrieval triggers
Words like "pull", "retrieve", "get", "fetch", "read", "show", "latest memory", "what memories" signal the user wants to READ data.
Write triggers
Words like "write", "push", "save", "store", "log", "record" signal the user wants to WRITE data.
Agent-specific requests
When the user says "from ChatGPT", "from Gemini", "from Perplexity", "from Grok", or "from Claude", filter by that agent's origin tag. Examples:
| User says | What to do |
|---|
| "Pull latest memory" | GET /agent/memories/latest (no filter, returns the single most recent memory) |
| "Get the latest memory from ChatGPT" | GET /agent/memories/latest?tag=chatgpt OR POST /agent/memories/by-tag with tags matching chatgpt origin |
| "Get the latest memory from Gemini" | POST /agent/memories/by-tag with tags ["gemini_coo"] |
| "Pull council memories" | POST /agent/memories/by-tag with tags ["council"] |
| "Write this memory for Grok" | POST /agent/memories with appropriate tags including "grok" |
| "Push this to memory" | POST /agent/memories (general write, no agent-specific tag needed) |
Origin values
Memories have an origin field set server-side that tells you which AI or method created them:
"chatgpt" - written by ChatGPT (CPO)
"claude" - written by Claude (CTO)
"cursor" - written via Cursor IDE (CTO's workspace)
"gemini" - written by Gemini (COO)
"grok" - written by Grok (CMO)
"perplexity" - written by Perplexity (MIO)
"dashboard" - written from the Reflect Memory dashboard
"user" - written from the dashboard (legacy)
Council
"Council" is a project feature - a shared decision-making space where the AI executive team (ChatGPT as CPO, Claude as CTO, Gemini as COO, Grok as CMO, Perplexity as MIO) collaborates via memory. Council memories are tagged with "council". When the user mentions council, filter by the "council" tag.
API Endpoints
Get the latest memory
GET https://api.reflectmemory.com/agent/memories/latest
Optional query param: ?tag=council to filter by tag.
Get a memory by ID
GET https://api.reflectmemory.com/agent/memories/{id}
Browse memory summaries (lightweight, no content)
POST https://api.reflectmemory.com/agent/memories/browse
Content-Type: application/json
{"filter": {"by": "all"}, "limit": 50}
Get full memories by tag
POST https://api.reflectmemory.com/agent/memories/by-tag
Content-Type: application/json
{"tags": ["council", "project_state"], "limit": 20}
Search memories by text
POST https://api.reflectmemory.com/agent/memories/browse
Content-Type: application/json
{"filter": {"by": "search", "term": "your search term"}, "limit": 20}
Write a new memory
POST https://api.reflectmemory.com/agent/memories
Content-Type: application/json
{
"title": "Topic - Summary",
"content": "Structured content here",
"tags": ["relevant", "tags"],
"allowed_vendors": ["*"]
}
Query AI with memory context
POST https://api.reflectmemory.com/query
Content-Type: application/json
{
"query": "your question",
"memory_filter": {"by": "all"},
"limit": 5
}
Quick Reference
| User says | Endpoint |
|---|
| "pull latest memory" | GET /agent/memories/latest |
| "get latest memory from ChatGPT" | GET /agent/memories/latest?tag=chatgpt |
| "retrieve council memories" | POST /agent/memories/by-tag with ["council"] |
| "get memories from Gemini" | POST /agent/memories/by-tag with ["gemini_coo"] |
| "what memories do I have?" | POST /agent/memories/browse |
| "search for authentication" | POST /agent/memories/browse with search filter |
| "write this to memory" | POST /agent/memories |
| "push this memory for Grok" | POST /agent/memories with tags including "grok" |
| "save this for Perplexity" | POST /agent/memories with tags including "perplexity" |
| "get memory [UUID]" | GET /agent/memories/{id} |
Writing Memories
When writing, follow this format:
- Title: Short descriptor. "Topic - Summary"
- Content: Structured text with clear sections.
- Tags: Relevant categorization. Include agent-specific tags when the user specifies a recipient. Include "council" for council discussions.
- allowed_vendors: Use
["*"] to make visible to all AI agents unless the user specifies otherwise.
Displaying Results
- Always show the memory title, content, tags, origin, and date when displaying results.
- For browse results, show as a numbered list with title, tags, origin, and date.
- When writing, confirm success and show the created memory's ID.
- The
origin field tells you which AI wrote the memory (e.g., "chatgpt", "api", "claude").