MCP tool reference

One start command.
Clear memory tools.

Run setup once and restart your AI tool. Engram then appears as MCP tools with startup guidance; you can inspect the focused tools below when needed.

1. Human action

Install once

You run one terminal command: npx engram-tools@0.3.6 --setup <api-key>. This writes MCP config for Claude Code, Cursor, Gemini, Antigravity, Codex, OpenCode, and VS Code/Copilot where possible.

$ npx -y engram-tools@0.3.6 --setup <api-key>
2. Agent action

Call tools inside the AI app

You do not run engram_start in your shell. After setup, the client exposes Engram tools and startup guidance to the agent. You describe the task normally; if a client ignores the guidance, paste a short context prompt once.

You: Work on this repo/task. Use Engram context before planning.
3. Why it works

MCP exposes Engram as tools

After restart, the AI client loads the Engram MCP server. The agent sees tools like engram_start, engram_store, and engram_recall, plus AGENTS.md guidance and Claude slash-command helpers.

Agent sees: engram_start, engram_store, engram_recall...

Troubleshooting config formats

Normal setup writes these automatically. Use this section only when a client does not show Engram tools after restart and you need to inspect the file shape.

Antigravity

Google's current Gemini flow may route users into Antigravity. Engram writes both the shared Gemini config and Antigravity-specific config files.

// ~/.gemini/config/mcp_config.json
// ~/.gemini/antigravity/mcp_config.json
{
  "mcpServers": {
    "engram": {
      "command": "npx",
      "args": ["-y", "engram-tools", "mcp"]
    }
  }
}

OpenCode

OpenCode uses a command array instead of the standard mcpServers shape. The setup command writes that format automatically.

// ~/.config/opencode/opencode.json
{
  "mcp": {
    "engram": {
      "type": "local",
      "command": ["npx", "-y", "engram-tools", "mcp"],
      "enabled": true
    }
  }
}

Codex

Codex uses TOML. Engram replaces only the engram table and preserves the rest of your Codex config.

# ~/.codex/config.toml
[mcp_servers.engram]
type = "stdio"
command = "npx"
args = ["-y", "engram-tools", "mcp"]

Tools

Select one tool to inspect

Core loop

The four tools agents should use in normal coding flow.

Advanced and checkpoints

Keep these available, but use them only when their trigger is real.

engram_start

Start a new agent session with one canonical call.

Call first after setup/restart, before the agent starts making product or code decisions.

agent_tool
optional
string

Which AI tool is calling, e.g. claude-code, codex, cursor, gemini, antigravity.

project
recommended
string

Product or client name, e.g. HouseCompass. The schema allows an empty start as a fallback, but useful starts should include this.

repo
recommended
string

Repository or project slug for filtering private memory. The schema allows an empty start as a fallback, but useful starts should include this.

role
optional
string

What the agent should behave as: frontend product engineer, backend engineer, PM, etc.

market
optional
string

Domain context, e.g. UK property, fintech, education.

task
recommended
string

Immediate task the agent is about to do. This is the strongest signal for relevant recall.

stack
optional
string[]

Relevant technologies.

context
optional
string

Optional full context summary. If omitted, Engram builds one from the other fields.

include_skills
optional
boolean

Default true. Include matching skill guides.

limit
optional
number

Default 12. Max memories to load.

Support Loop

Help us make Engram sharper

If setup is confusing, a tool behaves strangely, or a pricing limit feels wrong, send it here. The message lands directly with us.

Feedback

Tell us what is missing

Bugs, confusing setup steps, pricing questions, and product ideas go straight to the Engram team.

10 more characters needed0/10