Other clients + the Skills add-on

Two routes for hosts not covered by a dedicated guide:

1. MCP — the same SSE / stdio shim works for any MCP-compliant client
2. Agent Skills — a markdown bundle compatible with agentskills.io, for clients that prefer skills over MCP (or want both)

One-shot installer (try this first)

npx @azrtydxb/novamem-init

Detects 30+ AI hosts (Claude Code/Desktop, Cursor, Kilo Code, OpenCode, Codex CLI, Gemini CLI, Copilot, Cline, RooCode, Continue, Factory, Windsurf, Amazon Q, plus 16 skill-only hosts) and wires up whatever it finds — MCP config, skill bundle, slash commands. Idempotent. Skip ahead only if your host isn't in that list or you want to script it.

MCP — generic config

Direct SSE (recommended)

For any host that supports remote MCP over SSE:

{
  "mcpServers": {
    "novamem": {
      "type": "sse",
      "url": "http://localhost:7778/mcp/sse",
      "headers": { "Authorization": "Bearer nm_..." }
    }
  }
}

Confirmed working with: Claude Code, Claude Desktop, Cursor, Kilo Code, Goose, OpenCode, Continue (recent builds), Cline, Roo Code.

Stdio shim (legacy)

For hosts that only speak stdio MCP, the @azrtydxb/novamem-mcp shim bridges stdio↔HTTP:

{
  "mcpServers": {
    "novamem": {
      "command": "npx",
      "args": ["@azrtydxb/novamem-mcp"],
      "env": {
        "NOVAMEM_BASE_URL": "http://localhost:7778",
        "NOVAMEM_TOKEN": "nm_..."
      }
    }
  }
}

Same tool surface (15 tools — memory_* + project_*).

Agent Skills add-on

If your client supports the Agent Skills format — Goose, OpenCode, OpenHands, Junie, Roo Code, Factory, and a growing list — you can drop the bundled skill into the client's skills directory alongside or instead of the MCP server.

The skill teaches the agent when to call the tools and how to phrase saves; MCP exposes the tools themselves. With both, the skill loads at startup (~100 tokens for name + description), expands on demand, and the agent calls memory_* / project_* over MCP.

Layout

The bundle lives at skills/novamem/ and follows the agentskills.io spec:

skills/novamem/
├── SKILL.md                 # name, description, when-to-use rules
└── references/
    ├── search.md            # memory_search / recent / today / neighbors / stats
    ├── remember.md          # memory_remember / update / forget + worthiness gate
    └── projects.md          # all 7 project_* tools

Install

Most clients want a copy or symlink under <workspace>/skills/ or ~/.config/<agent>/skills/. Examples:

# Goose
cp -r skills/novamem ~/.config/goose/skills/

# OpenCode (project-scoped)
cp -r skills/novamem .opencode/skills/

# Generic — symlink works for any client that scans a skills dir
ln -s "$(pwd)/skills/novamem" ~/.skills/novamem

Check your client's docs for the exact path — the agentskills.io client list links each one.

Validate

npx -y skills-ref validate ./skills/novamem

This checks the SKILL.md frontmatter and naming conventions.

Build your own integration

For a host without dedicated docs:

1. Mint a nm_… bearer from the dashboard's API Tokens page.
2. Try the SSE config first — that works for any modern MCP host.
3. Fall back to the stdio shim if SSE isn't supported.
4. Optionally drop the Skills bundle in the client's skills directory for richer behaviour rules.
5. If your client speaks neither MCP nor Skills, use the HTTP API directly — the OpenAPI spec covers every operation.

If you wire up a new client, a PR adding it to this directory is welcome.