mirror of
https://github.com/agentscope-ai/ReMe.git
synced 2026-10-16 04:33:12 +00:00
6cb81e7921
* refactor(integrations): split TypeScript host plugins * feat(integrations): refresh host compatibility and status UI * fix(openclaw): secure status diagnostics
66 lines
2.9 KiB
Markdown
66 lines
2.9 KiB
Markdown
---
|
|
title: Agent Integrations
|
|
description: Connect ReMe to agents through the CLI, HTTP, MCP, Skills, and host adapters.
|
|
---
|
|
|
|
# Agent Integrations
|
|
|
|
ReMe keeps memory in an independent service and a user-owned workspace. Multiple agents can call the same memory system without binding storage to one model or host.
|
|
|
|
## Choose an interface
|
|
|
|
| Scenario | Recommended interface |
|
|
|---|---|
|
|
| Local script or hook | ReMe CLI |
|
|
| Application backend | HTTP Client |
|
|
| Tool-protocol host | MCP |
|
|
| DeepSeek Harness | `@agentscope-ai/reme-dsh-plugin` |
|
|
| OpenClaw | `@agentscope-ai/reme-openclaw-plugin` |
|
|
| Claude Code | MCP + Skill + Stop Hook |
|
|
| Hermes Agent | Memory provider adapter |
|
|
| Codex or another coding agent | `reme_memory` Skill or MCP |
|
|
|
|
## General memory loop
|
|
|
|
1. Before answering, call `search` for relevant memory.
|
|
2. Use `read` on high-value results and `traverse` when relationships matter.
|
|
3. Retain workspace-relative source paths in the answer.
|
|
4. At session end, pass source messages to `auto_memory`.
|
|
5. Let background or scheduled workflows consolidate daily notes into digest memory.
|
|
|
|
An empty search result must remain empty; do not present model inference as recalled history.
|
|
|
|
## MCP
|
|
|
|
The default HTTP service exposes streamable HTTP MCP at `http://127.0.0.1:2333/mcp`. Common tools include `search`,
|
|
`read`, `traverse`, `list`, `auto_memory`, and `proactive_read`.
|
|
|
|
Use `service.jobs` to expose a read-only subset or keep write tools in a separate configuration.
|
|
|
|
## CLI and Skill
|
|
|
|
`skills/reme_memory/SKILL.md` defines a general workflow for agents that can run local commands: installation checks, service discovery, retrieval, reading, and persistence boundaries.
|
|
|
|
It deliberately avoids silently modifying Python environments, stopping unknown processes on port conflicts, writing recalled tool output back as conversation source, or persisting credentials.
|
|
|
|
## TypeScript, OpenClaw, and DeepSeek Harness
|
|
|
|
Install the self-contained [DeepSeek Harness](./integrations/dsh.md) or [OpenClaw](./integrations/openclaw.md) plugin.
|
|
Each package owns its ReMe HTTP boundary and can evolve with its host independently.
|
|
|
|
## Claude Code
|
|
|
|
`integrations/claude_code/` provides streamable HTTP MCP configuration, a `reme-memory` Skill, and a Stop hook that calls `auto_memory_cc`. Follow that directory's README for installation.
|
|
|
|
## Hermes Agent
|
|
|
|
`integrations/hermes_agent/` provides a memory provider with HTTP and embedded modes. It recalls context before model calls and asynchronously invokes `auto_memory` after each turn. Its `config_schema.py` is rendered by Hermes' generic memory settings UI.
|
|
|
|
## Production guidance
|
|
|
|
- choose a stable absolute `workspace_dir`;
|
|
- reuse a service discovered by `reme find_reme`;
|
|
- treat `reme help` as the active Job contract;
|
|
- apply timeouts and failure logging to writes;
|
|
- do not block the host's core response path when memory is temporarily unavailable;
|
|
- use authentication, TLS, and a minimal Job allowlist for remote access.
|