@slashml/mcp
v1.1.3
Published
Official Model Context Protocol (MCP) server for SlashML (Claude Desktop, Claude Code, OpenAI Codex, Cursor, Windsurf, Cline, Zed)
Maintainers
Readme
@slashml/mcp -- SlashML MCP Server
Official Model Context Protocol (MCP) server for SlashML (/ ml). Connects coding and chat agents to the SlashML Cloud API (https://api.slashml.ai).
Quick Install
Claude Code (1-liner)
claude mcp add --scope user ml-memory -- env SLASHML_API_KEY=sml_live_xxx npx -y @slashml/mcp@latestCursor / Claude Desktop / Windsurf (mcp.json)
{
"mcpServers": {
"ml-memory": {
"command": "npx",
"args": ["-y", "@slashml/mcp@latest"],
"env": {
"SLASHML_API_KEY": "sml_live_xxx"
}
}
}
}Code Mode
Expose a single execute_memory_code meta-tool:
claude mcp add --scope user ml-memory-code -- env SLASHML_API_KEY=sml_live_xxx npx -y @slashml/mcp@latest --code-modeAmbient Lifecycle Hooks
Register UserPromptSubmit, PreToolUse, PostToolUse, and Stop hooks in ~/.claude/settings.json:
npx -y @slashml/mcp@latest --install-hooks --api-key "$SLASHML_API_KEY"Available Tools (Standard Mode)
search_memories-- Search stored conversational memories, user preferences, and facts.get_memories-- List conversational memories for a user/agent scope.forget_memory-- Delete or subject-purge (purge: true) a specific memory by ID.redact_memory-- Redact sensitive spans statelessly (text) or in-place (memory_id) under a policy preset (tier1_secrets,soc2_coding,gdpr,hipaa,enterprise_strict).dismiss_intervention-- Record that an active pre-tool intervention was a false positive or superseded for the current target.index_repository-- Index a git repository for code search and bug localization.search_code_by_behavior-- Search indexed repository symbols by behavioral or architectural description.find_bug_candidates-- Rank likely bug locations across a repository from symptoms and stack traces.add_memory(opt-in via--enable-write-tool; ambient lifecycle hooks assimilate memories automatically by default) -- Explicitly store a persistent conversational memory, user preference, or operational rule.
Per-Agent Access & Redaction Environment Variables
SLASHML_AGENT_ID(or--agent-id): AttachesX-SlashML-Agent-Idto enforce per-agentread_write/read_onlyaccess modes and redaction clearances configured inhttps://app.slashml.ai(Redaction & Access).SLASHML_REDACT_POLICY(or--redact-policy): Default redaction policy preset or comma-separated buckets.SLASHML_REDACT_SCOPE(or--redact-scope):memorize_and_recallorrecall_only.SLASHML_RUN_ID/SLASHML_PROJECT_ID(or--run/--project): Scope memories and guardrails to an active swarm run and project.
Swarm Guardrails & CLI Commands
# Wrap any stdio or HTTP MCP server with synchronous policy blocking & output redaction
npx -y @slashml/mcp wrap-mcp -- npx -y @modelcontextprotocol/server-postgres "$DATABASE_URL"
npx -y @slashml/mcp wrap-mcp --url https://mcp.example.com/mcp
# Install container PATH shims for mutating shell commands
npx -y @slashml/mcp install-shell-guards --bin-dir /usr/local/slashml-guards/bin
# Mint scoped worker tokens, promote validated run findings, or revoke tokens instantly
npx -y @slashml/mcp mint-token --agent worker-1 --tier worker --run run_42 --project proj_payments --print-env
npx -y @slashml/mcp promote --run run_42 --project proj_payments --discard-unpromoted
npx -y @slashml/mcp revoke-token --agent worker-1License
Apache-2.0. Connects to the SlashML Cloud Service (https://api.slashml.ai).
