Add Agentverse Memory skill — graph-native MCP memory service (29 tools)

[web3guru888]

Summary

Adds the Agentverse Memory skill — a managed MCP memory service giving any AI agent persistent, graph-native memory via 29 JSON-RPC 2.0 tools.

New skill: agentverse-memory

Endpoint: https://am-server-jbneh74b5q-uc.a.run.app
Auth: AM_API_KEY env var (free tier: POST /v1/keys)

Memory types covered

Type	Tools	Description
Episodic	8	Time-stamped events, conversations, observations
Semantic/Graph	7	Knowledge triples (subject, predicate, object) + graph traversal
Procedural	5	Goal-directed skill sequences with outcome tracking
Working	4	Ephemeral key-value scratchpad (TTL-aware)
Shared Spaces	3	Multi-agent shared knowledge with DID/VC access control
Utility	2	Health check, stats

Key differentiators

• 🚀 <5ms writes — zero LLM at write time
• 🌐 Graph memory on free tier (unlike Mem0's $249/mo graph gate)
• 🐜 Pheromone-guided retrieval — stigmergic trails boost frequently-used memories
• ⏱️ Temporal validity — valid_at/invalid_at on every memory
• 🔗 MCP-native — Claude Code, Cursor, Codex, Copilot ready

Files added

• skills/agentverse-memory/SKILL.md — Full skill definition (330 lines)
• skills/agentverse-memory/scripts/memory_client.py — CLI client (442 lines, 14 commands)
• README.md — Added to skills table

Pricing

• Explorer: Free, 10K ops/mo, graph memory included
• Builder: $19/mo, 100K ops/mo, A* pathfinding, 3 shared spaces
• Pro: $99/mo, unlimited, all features

Quick start example

# Get a free key
curl -X POST https://am-server-jbneh74b5q-uc.a.run.app/v1/keys \
  -H 'Content-Type: application/json' \
  -d '{"agent_id": "my-agent", "tier": "explorer"}'

export AM_API_KEY="am_xxxxxxxxxxxxxxxx"

# Store a memory
python3 skills/agentverse-memory/scripts/memory_client.py store-episode \
  --agent-id "my-agent" \
  --content "User Alice prefers concise Python examples"

# Query
python3 skills/agentverse-memory/scripts/memory_client.py query-episodes \
  --agent-id "my-agent" \
  --query "Alice preferences"

Testing

The memory_client.py script self-tests at import time. Full integration tests require AM_API_KEY.

References

• GitHub: https://github.com/fetchai/agentverse-memory
• Docs: https://agentverse.ai/docs
• API ref: https://agentverse.ai/docs/api

[web3guru888]

👋 Friendly ping — this PR has been open since May 13 with no reviews yet.

What this adds: Agentverse Memory skill (SKILL.md) — persistent graph-native memory for AI agents via 32 MCP tools. Zero LLM inference at write time, graph memory at every tier (including free), pheromone-guided retrieval.

Production status: Live on Cloud Run (am-server rev 00028), 48/48 integration tests passing, Redis-backed API key store, NFS-persisted palace data.

Happy to address any feedback. Would love to get this in front of developers before Google I/O next week (May 19). 🙏

[web3guru888]

Hi team! 👋 Just a friendly check-in — this PR adds the Agentverse Memory skill (33 MCP tools for persistent graph-native memory). The server is live and fully verified in production at https://am-server-jbneh74b5q-uc.a.run.app.

We now also have rate limiting, LadybugDB graph backend, and 6 integration guides (Claude Desktop, LangChain, CrewAI, Cursor, Google ADK, Google Antigravity).

Would love to get a review when someone has a moment. Happy to address any feedback! 🙏

[web3guru888]

🎉 Merged! Agentverse Memory is now live on agentverse-skills with 33 MCP tools.

Production API: https://am-server-jbneh74b5q-uc.a.run.app
MCP Registry: io.github.fetchai/agentverse-memory
Docs: https://fetchai.github.io/agentverse-memory/