Skip to content

CLI ​

engram <command> [options]
CommandWhat it does
initSets up .engram/, connects agents, installs skills, indexes the repo
serve (mcp)Runs the MCP server over stdio
statusCounts by kind and status, last activity
summary [--full]Project overview
context [task…]Prints the context pack an agent would get (-b budget, -f files, --raw)
remember <title> (add)Stores a memory (-k kind, -c content, -t tags, -f files, -i importance, -s status, --supersedes id)
recall [query…] (search)Searches (-k kinds, -t tags, -f files, -a include inactive, -n limit, --json)
show <id> (get)One memory with its links (--json)
link <from> <relation> <to>Connects memories (-n note)
graph <id>Shows the neighborhood in the graph (-d depth)
why <topic…>History of a topic
check <plan…>Checks a plan against pitfalls (-f files)
forget <id>Archives, or deletes with --hard
indexImports ADRs and git signals (--no-adr, --no-git)
export--format jsonl or md, -o file
import <file>Merges a JSONL export
syncRe-reads and cleanly rewrites memories.jsonl
skills listLists the bundled skills
skills installCopies skills into .claude/skills (--global, --dir)

init options ​

Option
-a, --agent <list>claude, cursor, agents-md, all. The default is to detect them.
-p, --project <name>Project name used in summaries
--no-skillsDon't install skills
--no-indexDon't scan ADRs and git history

Environment ​

Variable
ENGRAM_ROOTForces the project root

Released under the MIT License.