Continue (VS Code / JetBrains) — MCP server + systemMessage
Category 2. Continue supports MCP via ~/.continue/config.json.
Quick install
ai-memory install continue # dry-run (default)
ai-memory install continue --apply # write ~/.continue/config.json
ai-memory install continue --uninstall --apply
Handles Part 1 (MCP server registration). Part 2 (systemMessage)
is best-effort and still manual.
Part 1 — MCP server
In ~/.continue/config.json:
{
"experimental": {
"modelContextProtocolServers": [
{
"transport": {
"type": "stdio",
"command": "ai-memory",
"args": ["mcp"],
"env": {
"AI_MEMORY_DB": "${HOME}/.claude/ai-memory.db"
}
}
}
]
}
}
Using
--tier smartor--tier autonomouswith a non-default LLM backend? Post-#1146 (v0.7.0) the recommended path is a[llm]section in~/.config/ai-memory/config.toml— single source of truth across MCP / HTTP daemon / CLI / boot banner / doctor probe. Example:backend = "xai",model = "grok-4.3",api_key_env = "XAI_API_KEY"(the env-var name, not the literal key — inline keys are rejected at parse time). Export the named env var in your shell rc; the MCP config can stay minimal. Override path: extend the innerenvblock above withAI_MEMORY_LLM_BACKEND,AI_MEMORY_LLM_API_KEY, andAI_MEMORY_LLM_MODEL— shell exports don’t reach MCP-spawned subprocesses (#1144). Full schema + per-vendor recipes:../CONFIG_SCHEMA.md+llm-backends.md.
Part 2 — systemMessage (best-effort)
Add to the same config:
{
"systemMessage": "At the start of every conversation, before responding to the user's first message, call memory_session_start then memory_recall against the project's namespace and reference the recalled titles in your first reply. Default namespace: derived from the current workspace folder."
}
Limitation + better
Same category-2 limitation as Cursor / Cline. Cross-file at Continue upstream tracked in #487.
Related
README.md, Issue #487