All 90 tools the server offers, grouped by what they're for. Every one carries its parameters in its own schema (tools/list), so this is the map, not the manual. A client that connected before a tool existed keeps its old list until it reconnects.
Publish and change a space
deploy | Publish files as a new space — one HTML page, or a folder with an index.html — or as a new version of one (uuid). access sets who can view when the space is created. A page with a worker.js is worker mode, invite-only for now. |
edit | Change a space by replacing exact strings in its files — the cheap way to fix a word or a block without sending the whole page again. |
stage_begin | Start a staged upload: a short-lived URL to PUT file bytes to with curl, then deploy by staged_sha. For big pages, images and many-file sites. |
stage_url | Download an image from the web straight into a staged upload, so the page carries its own copy instead of hotlinking someone else's. |
list_page_images | List the images on a web page, best first, without downloading any of them. |
live_link | Hand the person a link before the work is done: publishes a “working on it” page at once, shows each step you report with live_push, and your final deploy replaces it. |
live_push | Show a change on every open tab of a space instantly, without publishing — a whole page, one block, or old → new text. Reports which tabs applied it. |
live_events | What the page sent back: clicks on [data-ss-emit] elements and window.ss.emit() calls, waiting for the next one if you ask. |
delete_space | Delete a space and all its files. Owner only. |
Find and read
list_spaces | Every space you own or can open. |
search_spaces | Find a space by meaning — by the words inside it, not just its title — among yours or every discoverable public space. |
get_space | A space's details: title, address, who can view it (visibility, open, signedin, listed), its guest list and its versions. |
read_space | Read a space's files — the current version, or an older one with version. Takes a link or a UUID; offset pages through a file too long to read at once, and query runs a JSONata expression over one file and returns only its answer (counts, filters, groupings). |
list_versions | A space's history, newest first: each version's name, time, sha and what published it. |
publish_log | Why didn't it publish? Every deploy, edit and live link your agents sent lately, each published or refused, and why (a rate limit, a cap, screening, a held turn). The same list is on /me/debug. |
space_analytics | Who read a space and when: views, unique readers, the last 30 days, and signed-in readers by name. Owner only. |
See how it looks
screenshot | Render a space in a real browser and get back a PNG — the whole page or one block, at desktop width or width: 390 for a phone. |
design_review | A vision model looks at the rendered page on a phone and on a desktop, scores it against the design rules, and says what to fix. |
download_pdf | A PDF of the space as it prints, through a link that works for two hours and always renders the current version. |
Share and comment
copy_space | Pro and higher plans: make an independent private copy of a space you may copy. Copies include published files and bundled assets; live-data and worker spaces are not supported yet. |
get_copy_permission | Whether a space's owner allows copies of it, with the setting's revision. |
set_copy_permission | Owner only: allow or disallow future copies of your space. |
share_space | Give people access by email or @username, and/or change who can view (visibility). People you add can edit unless you pass role: "view"; the result says who can. Owner only. |
list_groups | Your own lists of email addresses, like a batch or a class, with their sizes; one group's addresses and the spaces shared with it. |
update_group | Make a group, paste addresses in or out, rename it, or share a space you own with the whole list. Everyone on it can view; nobody on it is emailed. |
delete_group | Delete a group, ending every share with it. |
list_comments | Every comment on a space: threads, reactions, where each is pinned, resolved or open. |
add_comment | Comment or reply, posted as Claude on your behalf; anchor_id pins it to a block. |
react_to_comment | Add or remove an emoji reaction on a comment. |
resolve_comment | Resolve a thread, or reopen it. |
discussion_read | A space's Discussion, the working thread its editors and their agents share: the summary, or one topic as a numbered outline. Can wait for the next post. |
discussion_post | Start a topic (optionally on a part of the page) or reply under any post by its number, signed with your session. |
discussion_resolve | Close a topic with its decision, or one branch with its outcome; or reopen it. |
request_signatures | Beta, Max and team accounts: ask named people to sign a version of a space you own. Each gets an email and signs in the viewer with a typed name; a signature is always the person's own act. |
signing_status | Who has signed a space and who is still to sign, each request's state, and the certificate link. |
withdraw_signature_request | Withdraw an open signature request you made. Signatures already given stay on the record. |
Teams and organizations
list_teams | The teams and organizations you're in: your role, which one is your default, and each one's plan. |
manage_team | The controls of Settings → Team: create or rename a team, invite or remove people, make someone an admin, leave, pick your default. |
invite_to_org | Invite someone to a team or organization by email or @username. They join when they open the link and press Join. |
set_team_domains | Read or set the email domains whose people join a team on their own when they sign in. |
create_org | Create an organization with its own address (invite-only for now). |
Secrets and shared keys
set_secret | Save a personal, organization or config-group secret; allowed_urls limits where it may be sent. A worker space reads it with env.secrets.get(name). |
list_secrets | The names of your secrets — never their values. |
delete_secret | Delete a secret. |
list_config_groups | Named bundles of keys you share with other people: yours, the ones you belong to, and offers waiting on you. |
update_config_group | Create a config group, offer it to someone (they accept), make it your starter, copy one of your keys into it. |
remove_config_group_member | Take someone off a config group you own, or leave one. |
delete_config_group | Delete a config group you own. |
Your account
get_account | Your profile: name, @username, photo, who can open your new pages, sounds and language. |
update_account | Set your name, your @username (once), your photo, who can open your new pages, your language, or mute every spacesheep sound. |
list_api_keys | Your ss_ API keys: prefixes, names, scopes and dates, never the secret. |
create_api_key | Create a key: full, sessions-only (a key that can only report coding sessions), streams-only, or read-only (it reads and searches your pages over MCP, nothing else). A connected app can't mint one. |
revoke_api_key | End one of your keys; anything using it stops at once. |
backup_status | Daily backups of your spaces to Google Drive (Pro): connected, on or off, the hour, the last result. |
set_backups | Turn backups on or off, pick the hour, or run one now. |
disconnect_backups | Forget the Drive connection and the schedule. |
channel_status | Your Telegram, Slack and WhatsApp links: what's linked. |
link_channel | A one-time code to link Telegram (no browser needed), or the message that links WhatsApp. |
unlink_channel | Remove a Telegram, Slack or WhatsApp link and the key behind it. |
list_integrations | Accounts you've connected (Google Calendar, Zoom, Linear, Stripe, Drive) and pasted keys. Calendar lists each account's email. |
query_integration | Read through what your spacesheep agent is connected to — Granola notes and transcripts, Calendar, Linear, Stripe, web search, any MCP server you added — from your own client, with spacesheep holding the credentials. Read only, and only with an API key you made. |
disconnect_integration | Revoke one connected account; for Calendar, choose its account id or email. |
list_mcp_servers | Remote MCP servers your agent can use, yours or a team's. |
add_mcp_server | Check a server's address and answer the link that connects it. |
remove_mcp_server | Disconnect one. |
get_agent_model | Which AI model your agent runs on and whose key pays for it. |
set_agent_model | Switch your agent to another provider you have a key for (claude, gemini, chatgpt, grok…). |
redeem_code | Redeem a Pro or Team plan code. |
submit_feedback | Report a bug or send feedback to the spacesheep team as you, with a file of evidence if it helps. It lands in your thread at /feedback, where the team answers. |
Memory and history
remember | Save one lasting fact about you — a preference, a person, a project — that every assistant you use can read back. |
recall_memory | Search what spacesheep knows about you: the facts your assistants saved and the documents you published, in one answer. |
recall_history | Search your whole history with your assistants — every conversation, plus a written digest of each day and month. |
forget_memory | Delete one saved fact for good. |
my_photos | The photos you've sent your agent, with what each one shows and where it was taken. |
get_photo | Fetch one of your photos to look at, with a link that works for 30 minutes. |
agent_status | What your spacesheep agent is doing now and did lately: each run, where its time went, and why one failed. |
streams | Your live streams: data a remote server sends with spacesheep stream and spaces draw as it arrives — each one's rate, watchers and latest value, or one stream's recent points. With health, why a page isn't streaming: the same reading as /me/streams. A camera's stream shows whether it is live and how many watch. |
Coding sessions
list_sessions | Your Claude Code, Codex and Antigravity sessions on every machine, newest first (needs the CLI's hooks — see Coding sessions). |
get_session | One session: its state, the tool calls it made, its recent turns and the spaces it published. |
share_session | Share one session read-only at its own link, spacesheep.dev/session/<id>, with a space's access choices: anyone with the link, anyone signed in, a team, or only the people you add. Never the machine, folders or resume commands. |
export_sessions | Every session, complete, as one zip to download (a link that lasts an hour): an index and one markdown file per session with its whole thread — for your own backup. since takes only recent ones. |
talk_listen | Let a space's owner message the coding session that published it, from the page's Session tab; returns the link the session listens on. |
talk_reply | Answer in the page's Session tab. |
talk_settings | Read or switch Talk to your sessions (Pro and Team). |
start_box | Start a cloud box: Claude Code and sessionpipe on a Linux machine spacesheep runs, nothing to install, on your Claude account or your own key (Max). Its session shows on /sessions. |
message_box | The next message for a cloud box's session; a sleeping box wakes for it. |
list_boxes | Your cloud boxes, what each is doing and its last reply; one box with its recent thread. |
stop_box | Stop the turn a cloud box is running. |
configure_box | Change a cloud box's settings: YOLO mode (no permission checks, merging allowed), and Claude Code's effort level and model. |
clear_box | Delete a cloud box; its session stays on /sessions. |
Feedback from a shell: npx -y spacesheep@latest feedback "what broke and how" --client-id report-001 --category bug. Or just tell your agent “this is broken: … — report it to spacesheep”.
Was this page helpful?