Skip to content

Dimension 4: Opencode Compatibility Analysis

Opencode's Memory Ecosystem

OpenCode (https://opencode.ai) is an AI coding agent platform that fully supports MCP (Model Context Protocol) integration. It provides structured tool calling, message history handling, and model orchestration.

MCP Server Support

OpenCode supports both local and remote MCP servers. Once added, MCP tools are automatically available to the LLM alongside built-in tools.

Memory Plugins Available for OpenCode

Plugin Provider Type Description
opencode-supermemory Supermemory Plugin Persistent memory across sessions using Supermemory cloud
honcho-opencode Honcho Plugin Persistent memory surviving context wipes and session restarts
mcp-memory-service Community MCP Server Open-source persistent memory with knowledge graph
Mem0 MCP Mem0 MCP Server Cloud-hosted MCP server at mcp.mem0.ai
Hindsight MCP Hindsight MCP Server OAuth-secured MCP at api.hindsight.vectorize.io/mcp
Super-Memory-TS Veedubin MCP Server Local-first semantic memory with Qdrant

How Memory Works in OpenCode

  1. MCP tools exposed to the LLM (e.g., memory_save, memory_search, memory_update)
  2. Agent decides when to call — the LLM uses tools to save/recall memories
  3. Context injection — on session start, relevant memories fetched and injected
  4. Smart compaction — at 80% context capacity, sessions summarized and saved
  5. Privacy — <private> tags redact content before storage

Supermemory's Deep OpenCode Integration

  • /supermemory-init command — deep research session exploring project
  • Three things injected on session start: user profile, project memories, semantic search results
  • Preemptive compaction at 80% context usage
  • Memory scopes: user (cross-project) and project (current directory)
  • Memory types: project-config, architecture, error-solution, preference, learned-pattern, conversation

Hermes Memory Provider Protocol vs OpenCode

The Core Incompatibility

Hermes memory provider protocol is NOT directly compatible with OpenCode. They are fundamentally different architectures:

Aspect Hermes Protocol OpenCode/MCP
Integration mechanism Python ABC class + plugin.yaml MCP (Model Context Protocol)
Language Python only Language-agnostic (JSON-RPC)
Tool exposure Provider-specific tools auto-registered MCP tools exposed via stdio/HTTP/SSE
Context injection Automatic — prefetch + sync_turn hooks Via MCP tool calls
Activation memory.provider in config.yaml MCP server config
Memory extraction Automatic session-end extraction Agent-initiated via tools
Multi-provider Single-select (one at a time) Multiple MCP servers simultaneously

Bridge Path: MCP Wrappers

Each Hermes provider that offers an MCP server CAN work with OpenCode: - Mem0 MCP: https://mcp.mem0.ai/mcp — works with OpenCode - Honcho MCP: honcho-ai/opencode-honcho package — works with OpenCode - Hindsight MCP: api.hindsight.vectorize.io/mcp — OAuth-secured, works with OpenCode - Supermemory: opencode-supermemory plugin — deep native integration

Super-Memory-TS Compatibility

Super-Memory-TS IS fully compatible with OpenCode because it uses MCP: - Runs as MCP server (stdio or HTTP) - 5 standard MCP tools exposed - Powers Boomerang-v2 (OpenCode plugin) - No translation layer needed

Key Insight for the User

The user's Super-Memory-TS project is architecturally aligned with OpenCode (both MCP-based). To incorporate ideas from Hermes providers, the user would need to either: 1. Add MCP wrappers around Hermes provider concepts 2. Port Hermes provider features into Super-Memory-TS directly 3. Build a new memory system that combines MCP compatibility with Hermes-style automatic extraction