Command line

Command
interlatch install [--yes] [--analyze all\|N\|later] [--[no-]notify-updates] [--[no-]menu-bar] [--dry-run]	Setup: lists the agents found, asks which to record, imports them, offers to analyze them now with progress (which builds the Glossary and the Map), starts the dashboard, and on macOS asks once whether to show the menu-bar icon (off unless you say yes) (Install)
interlatch app	The desktop app (window + menu bar); needs the app extra
interlatch ui [--open] [--menu-bar]	Dashboard (also always running at :11524 after install; if it is, interlatch ui says so and --open opens it). On macOS the one that runs at login also shows the menu-bar icon if you turned it on at install; --menu-bar/--no-menu-bar overrides that. Sessions, Knowledge, Projects and Glossary switch between Cards and List (a sortable table; click a row for details), remembered per page
interlatch sessions [-p project] [--since 7d]	List sessions
interlatch show <id-prefix> [--transcript\|--markdown\|--json]	Session overview or full conversation
interlatch search <words>	Full-text search over transcripts + knowledge (any language, 3+ chars)
interlatch knowledge [query] [-k gotcha] [-p project]	Browse extracted knowledge
interlatch projects / interlatch stats [--since 30d]	Per-project and overall statistics
interlatch analyze <id> \| --pending [--limit N] [--dry-run] [--backend codex]	Analyze now (--dry-run shows digest sizes, no tokens spent; --backend picks the agent or model provider for this run only)
interlatch synthesize [--project P] [--global] [--all]	Rebuild knowledge bases
interlatch export [--full]	Rewrite the Markdown vault
interlatch export <id>… [--format md\|json\|raw] [--out PATH]	Export sessions: one file, or a .zip of several (raw: the original transcript)
interlatch glossary [term] [-p project] [--rebuild --all] [--themes]	Your vocabulary: internal names, acronyms, domain terms with definitions and usage; --themes groups big categories into themes for the Map
interlatch systems [NAME] [--links] [--evidence] [--json]	The Systems map: every project as a system, grouped by folder, with what it runs on; NAME for one system's parts, connections and evidence (--evidence: with example commands)
interlatch review [2026-W39\|current]	Weekly engineering review written by the analysis model (automatic for each completed week)
interlatch import <zip> [--analyze] [--screen]	Import chats from a claude.ai or ChatGPT data export (the .zip, its folder, or conversations.json); repeatable. --screen screens them right after. See Sources
interlatch screen [--source chatgpt\|claude-ai] [--sample N] [--dry-run] [--redo]	Sort imported chats into worth analyzing, maybe and not worth it, reading only each chat's opening (Screening imported chats)
interlatch screen --list analyze\|maybe\|skip [--json] / --queue [--maybe]	Show the chats screened as one verdict, with why; queue the ones worth analyzing (and the maybes) for the background analysis
interlatch friction [-p project] [--days N] [--json] [--noise]	What goes wrong: recurring failure causes across your sessions (sessions, projects, last seen, still happening, trend), then the tools that fail most; --noise also lists expected failures (Suggestions)
interlatch suggest [-p project] [--all] [--json]	Proposed fixes waiting for you: instruction lines, config changes, setup steps. Nothing is written until you apply one (Suggestions)
interlatch suggest show\|apply\|dismiss\|done\|undo ID… [--yes] [--reason R]	Show the diff; apply it (asks first unless --yes, backs the file up); never propose it again; mark a setup step you ran as done; take an applied one back out
interlatch suggest move ID --to user\|project	Put a waiting line in your user-level file, or back in the files of the projects it came from; remembered for that lesson or cause (Moving a line)
interlatch suggest refresh	Look at the latest sessions and knowledge for suggestions now (the background sync does this after each run)
interlatch forget <id> [--delete-transcript]	Remove a session from the vault for good (it is never re-ingested)
interlatch sources	Which agents are connected, and how
interlatch connect <agent> / disconnect <agent>	Start/stop recording claude, codex, codex-cloud, copilot, bob or antigravity (data is kept)
interlatch connect <client> / disconnect <client>	Add or remove the MCP server in claude-desktop, cursor, windsurf or gemini
interlatch mcp [--print-config]	Run the MCP server (clients start it), or print a config entry for any other MCP client
interlatch status	Health: hooks, agents, MCP, queue, failures
interlatch migrate [--dry-run]	Move a Chronicle setup over: ~/.claude-chronicle to ~/.interlatch, and hooks, MCP servers, permission rules and background agents to the new names. Runs by itself when the dashboard, the sync, install or the app starts; --dry-run shows what it would change (Moving from Chronicle)
interlatch tailnet on\|off\|status [--anyone]	Open the dashboard on your phone and other devices through Tailscale Serve, to your Tailscale login only (--anyone: everyone in your tailnet) (Phone and other computers)
interlatch mirror [status] / mirror sync [--full]	The copy of this archive in Postgres: where it is, what each table holds and the last write; or write it now (--full: every row again) (A copy in Postgres)
interlatch hub enable [--rotate]	Make this computer the hub for your others; prints the interlatch hub join command to run on them
interlatch hub join <address> --token <token> \| --code <code> [--share knowledge] [--all-folders] / hub leave	Send this computer's sessions to a hub instead of recording them here, or stop (Your other computers). --code: the invite code an admin gave you (Joining a computer). --share knowledge: keep recording and analyzing here and send only summaries and project lessons (Sharing knowledge only); always the case for someone limited to projects. Sharing knowledge, it shares only sessions in the hub's projects unless --all-folders
interlatch hub status / hub disable	The computers sending to this hub; stop accepting them
interlatch hub project add <folder> / hub project remove <folder> / hub project list	On the hub: set up a project before anyone sent to it (that folder on the hub computer and everything below it, named after the folder), undo one, or list the hub's projects with the ones set up here marked (Projects and who sees them)
interlatch hub invite <name> [--email <email>] [--role admin\|member\|readonly] --project <name>… \| --all-projects	On the hub: add a person and make a one-time invite code (Inviting someone). --project (repeat it for more) or --all-projects: the projects a member or read-only person sees; a new one needs one of the two, an admin sees every project. Run again for someone already on the hub: a new code. --computer <name or id>: a code for that one computer the hub already knows, when it can't show its key (Joining a computer)
interlatch hub access <email\|id> --project <name>… \| --all-projects	On the hub: change which projects someone sees, from their next request. It replaces the list: name every project they keep (Projects and who sees them)
interlatch hub purge <email\|id\|computer> --project <name>… \| --outside-access [--yes]	On the hub: remove for good what a person's computers (or one computer) sent, in those projects or outside the ones the person sees (Taking back what a computer sent)
interlatch hub people / hub role <email\|id> <role> / hub remove <email\|id> / hub shared-token on\|off	On the hub: everyone with their role, the projects they see and their computers; change a role; remove someone; allow or refuse the shared token (People and roles)
interlatch hub add-folder <folder> --project <name> / hub remove-folder <folder> / hub folders [--list]	On a computer that joined a hub: file a folder's sessions under a project on the hub, take a folder added by mistake back out, or show what goes where (--list: the hub's projects) (Same project, different folders). Sharing knowledge, remove-folder also has the hub delete what the computer shared from that folder (Leaving)
interlatch hub leave --project <name> / hub rejoin --project <name>	On a computer that shares knowledge: stop sharing to one of the hub's projects and getting its teammates' lessons, keeping what it already shared there; or start again (Leaving)
interlatch push	On a computer that joined a hub: send its new sessions now (the hook and the background sync do this)
interlatch container	What the hub's Docker image runs: set the hub up from INTERLATCH_* variables, then serve the dashboard and sync every 15 minutes (A hub in Docker)
interlatch config [edit]	Show or edit ~/.interlatch/config.toml
interlatch config set <section.key> <value>	Change one setting, e.g. interlatch config set analysis.backend codex or providers.ollama.model qwen3:30b (Configuration)
interlatch config set-key <provider> [KEY] / forget-key <provider>	Store a model provider's API key, or IBM Bob's (bob), in provider-keys.json (asked for when left out), or remove it

Other ways in

Markdown vault: ~/.interlatch/notes (open it as an Obsidian vault): Home.md, Sessions/YYYY/MM/*.md with YAML frontmatter, Projects/*.md (knowledge base + session list), Knowledge/<Kind>.md, Reviews/YYYY-Www.md, Glossary.md, Global Playbook.md.

Inside your agents (MCP tools, registered in Claude Code and in each connected agent): search_knowledge, search_sessions, get_session, get_transcript, project_knowledge, glossary, find_artifacts, recent_sessions. Ask e.g. "have we hit this error before?" or "what is the deployer_ip rule?".

The MCP server is registered when you connect an agent (see Sources). MCP server covers each tool and how to connect other clients.