About memoryoracle
MemoryOracle gives AI agents persistent memory across sessions, with categorized storage, full-text search, and session summaries. It also provides namespace isolation, memory version history, and deletion and export controls.
Key features
- 01Persistent storage and full-text search
- 02Manage memories by category and namespace
- 03Summarize sessions and extract facts and tasks
- 04Track memory version history
- 05Delete and export memories
Use cases
- 01Retain user preferences and project decisions across sessions
- 02Retrieve past conclusions from compliance checks
- 03Organize session summaries, facts, and action items
MCP support
MemoryOracle exposes 10 memory-management tools over MCP for clients to store, retrieve, update, delete, and export memories.
Configuration
Remote servers need no local install. Pick your client, then copy the config or install in one click.
https://tooloracle.io/memory/mcp/claude mcp add --transport http io-tooloracle-memoryoracle https://tooloracle.io/memory/mcp/{"mcpServers":{"io-tooloracle-memoryoracle":{"type":"http","url":"https://tooloracle.io/memory/mcp/"}}}Open in Cursor{"servers":{"io-tooloracle-memoryoracle":{"type":"http","url":"https://tooloracle.io/memory/mcp/"}}}Open in VS Code{"mcpServers":{"io-tooloracle-memoryoracle":{"type":"http","url":"https://tooloracle.io/memory/mcp/"}}}{"mcpServers":{"io-tooloracle-memoryoracle":{"serverUrl":"https://tooloracle.io/memory/mcp/"}}}Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.
Tools
Details
From the project READMEMemoryOracle
Persistent long-term memory for AI agents. Agents never forget.
10 tools. SQLite FTS5 full-text search. Per-namespace isolation. GDPR-compliant forget. Cross-references with FeedOracle Trust Layer.
npx -y mcp-remote https://mcp.feedoracle.io/memory/mcp/
The problem
Every MCP tool call is stateless. When Claude Desktop restarts, when Cursor closes, when the session ends — the agent forgets everything. No user preferences, no project history, no learned rules, no portfolio tracking.
MemoryOracle gives agents a persistent, searchable memory that survives restarts, works across clients, and gets more valuable over time.
Tools
| Tool | What it does |
|---|---|
store_memory | Store a fact, preference, decision, or rule. Deduplicates automatically. |
query_memory | Full-text search with BM25 ranking and recency boost |
update_fact | Update a memory with new info. Preserves version history (last 10 changes) |
forget | Permanently delete by ID, category, age, or everything. GDPR Art. 17 |
summarize_session | End-of-session capture: summary + extracted facts, decisions, action items |
list_memories | Browse with filters: category, importance, recency, access frequency |
cross_reference | Verify a memory against FeedOracle Trust Layer evidence |
memory_stats | Dashboard: count, storage, categories, most accessed, recent |
export_memories | Export all as JSON. GDPR Art. 20 data portability |
health_check | Service status |
Memory categories
| Category | When to use |
|---|---|
fact | Verified information: "USDC is MiCA authorized" |
preference | User/agent preferences: "Conservative allocation, no USDT" |
decision | Decisions made: "Blocked USDT for EU clients" |
observation | Patterns noticed: "RLUSD peg stable for 30 days" |
rule | Business rules: "Never recommend assets with peg deviation > 2%" |
session | Session summaries with linked facts |
portfolio | Portfolio positions and allocations |
action_item | Open tasks: "Check RLUSD status in April" |
Example: Compliance agent workflow
Session 1 — Agent does MiCA compliance checks:
{"name": "store_memory", "arguments": {
"content": "USDT blocked for EU — MiCA NOT AUTHORIZED. Preflight: STOP.",
"category": "decision", "importance": 10,
"tags": ["usdt", "mica", "blocked"],
"namespace": "compliance-bot"
}}
Session 2 — Next day, agent remembers:
{"name": "query_memory", "arguments": {
"query": "USDT EU compliance status",
"namespace": "compliance-bot"
}}
→ Instantly recalls the decision from yesterday. No re-checking needed.
End of session — Auto-capture:
{"name": "summarize_session", "arguments": {
"summary": "Reviewed stablecoin portfolio. USDT blocked. RLUSD pending.",
"key_facts": ["USDC authorized", "EURC authorized", "USDT blocked"],
"decisions": ["Block USDT for EU"],
"action_items": ["Check RLUSD in April 2026"],
"namespace": "compliance-bot"
}}
Why this is different
| Feature | MemoryOracle | Others |
|---|---|---|
| Search | FTS5 BM25 ranking + recency boost | Basic keyword |
| Categories | 8 structured types | Flat key-value |
| Session capture | Auto-extracts facts, decisions, actions | Manual only |
| Trust Layer | Cross-reference with signed evidence | None |
| GDPR | Art. 17 forget + Art. 20 export | Usually missing |
| Namespaces | Per-agent isolation | Shared global |
| History | Version tracking (last 10 changes) | Overwrite |
| TTL | Auto-expiry with configurable hours | None |
x402 Pay-per-call
POST https://tooloracle.io/x402/memory/mcp/
| Tool | Units | Price |
|---|---|---|
store_memory | 2 | $0.02 |
query_memory | 5 | $0.05 |
summarize_session | 8 | $0.08 |
cross_reference | 5 | $0.05 |
list_memories | 3 | $0.03 |
update_fact | 2 | $0.02 |
forget | 1 | $0.01 |
Connect
{
"mcpServers": {
"memory": {
"command": "npx",
"args": ["-y", "mcp-remote", "https://mcp.feedoracle.io/memory/mcp/"]
}
}
}
Part of ToolOracle
MemoryOracle is part of ToolOracle — 50+ MCP servers, 530+ tools. Powered by FeedOracle compliance infrastructure.
The memory layer that makes Decision Preflight, Trust Layer, ISO 20022, and every other tool exponentially more valuable.
Read More
📝 We Built an AI Agent That Runs 24/7, Never Forgets, and Checks Its Own Work — How Memory + Scheduler + Decision Preflight work together as one integrated autonomous agent stack. Real tool outputs, no theory.
License
MIT
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.