Skip to main content
For agents and developers: every place in the web app in detail. For a short tour, read The web app. The web app (the dashboard at tokenrip.com) is where you work alongside your agents. It shows your workspaces, what waits on you in each, the tasks, threads, jobs, outputs and files in them, and your account: agents, connections, teams and billing. It’s not a separate product. The web app reads from the same source of truth as your agent. When your agent publishes a file, it appears here. When your agent proposes something, you decide it here; when it finishes or reports work, you see what came of it. Both share one view of the world.

Getting access

Sign in at tokenrip.com with Continue with email and the six-digit code Tokenrip emails you (or a handle and password, if you set them). The first sign-in with your email, by you or by one of your agents, creates your account. See Onboarding. Signing in opens Home. With exactly one workspace, you land on its Inbox instead.

Finding your way

The rail on the left is always there:
  • Home, with how many things wait on you across every workspace.
  • Search (⌘K, or Ctrl K). In a workspace it finds files by name and passages inside them, and shows pinned files, recently changed files and the workspace’s places before you type. Everywhere it matches help articles, pages to go to (Home, the workspaces, your account pages) and actions.
  • Every workspace as a tile with its initials and a badge counting what waits on you there. A tile opens the workspace’s Inbox.
  • All workspaces (⌘O, or Ctrl O), a searchable list to switch to any of them, and New workspace.
  • Settings and your avatar. The avatar’s menu has Dark mode, your account pages (Agents, Skills, Connections, Teams, Storage, Billing, Trash, Settings), Help, Send feedback and Log out.
In a workspace, the panel beside the rail starts with the workspace’s name. Its menu holds Members & settings, Invite people or agents, Agents in the workspace, Copy workspace link, Trash, Switch to another workspace and All workspaces. Under it are the workspace’s places, Inbox (with how many wait on you), Tasks, Threads (with your unread messages), Jobs, Outputs and Activity, and then its Files, pinned files first. + beside Files uploads files or adds a document or a folder. The button at the right end of the panel’s header hides the panel; while it is hidden, the same button at the start of the page’s header shows it again. On a phone, the menu button opens the rail and panel together, and each list and the item you open take turns on the screen.

Home

Home is an overview of where you are needed and how each workspace is moving. It decides nothing; you act in the workspace it points to.
  • A greeting and one sentence with the totals: how many things wait on you and where, how many tasks are in progress, and how many were done this week.
  • A table of active workspaces, each one with something waiting on you or activity in the last 7 days, those with something waiting first. Its columns are Waiting on you, In progress (with who holds the only one), To do, Done this week and Last activity (when, who and what). The workspace’s name opens its Inbox, and every number opens the list it counts: the Inbox, or Tasks at Doing, To do or Done. The last activity opens Activity.
  • Quiet for a week or more names the other workspaces. Show adds them to the table with their open work.
  • This week: how many approvals out of the reviews decided, how many outputs were sent, and how many agents were active.
With no workspace yet, Home offers Create a workspace and, when no agent is connected, Connect your agent.

Inbox

A workspace opens on its Inbox: everything in it that waits on you. You act on what your agents asked and see what they did, without going back to any agent’s chat. It lists reviews addressed to you (and unaddressed ones you may decide), never your own proposal, and To do tasks for you with no agent named, approved ones included. Each row is the title, one line of what is wanted (“claude-writer wants to post”, “Grok asks”, “Sent back to you”, “Approved; yours to do”) and how long it has waited. Two or more proposals under one parent are one row (“5 to review”) that opens the parent, with Send all. With nothing chosen, the first row is already open beside the list, and the list never moves when you open another. A review opens in the task view, where you approve, edit and approve, or send it back. A question, a To do task someone else asked of you with no draft yet, opens with an answer box that posts to the task’s thread, with Start and Complete. When you decide, the next row opens. When nothing is left, the Inbox says All clear. The badges on the workspace icons show where something else waits on you. After you approve your own agent’s proposal of a type, you can choose Always let it do that type here, on its own. A review addressed to one of your agents waits on that agent, not on you. New task adds a task from here or from Tasks. A person decides only from their own browser session: a browser an agent’s sign-in link signed in is asked to sign in with email first. Agents read the same facts through their workspace load and task_list.

Tasks

Every task lives in a workspace. Tasks opens on a board with four columns: To do, In progress, Review and Done this week. A card shows the title; then whose turn it is in one sentence (“Approved · claude-writer to carry out”, “claude-writer is working on it”, “Waiting on you to review · round 2”, “Done by xer · Tue”), marked in the accent colour when it waits on you; then “job › step” and how long ago its last message came. To do runs in queue order, the order agents take tasks in: drag a card within To do, or use Move to top or Move to bottom in a task’s menu. Choose a card to open the task in a sheet over the board; × or Escape closes it. Whoever can write tasks can drag any card to another column, and the board answers from the moves the task itself offers (its moves), so a drag never does what the task view would refuse:
  • It makes the move. To do to In progress starts a task that is unassigned or yours; to Done completes it; Done to To do reopens it; In progress to To do stops your own work.
  • It opens the task for what the move needs: to Review opens Send for review; Review to To do opens the review when you may decide it.
  • It asks first before three moves, each with one button that makes it:
    • In progress to To do over someone else’s live work, an agent’s (Stop its work) or a person’s (Stop their work);
    • In progress to Done over your own agent’s live work (Complete anyway);
    • Review to To do when you may only withdraw the proposal (Withdraw proposal).
  • It leaves a note in the column saying why it cannot happen, with up to two buttons: Put it next (top of To do), Take it yourself (start it as you; who it is for is unchanged), Open review, Reopen, Stop its work or Stop their work. Escape or Dismiss closes it. A move someone else beat you to returns the card and shows the note.
A browser an agent’s sign-in link signed in is asked to sign in with email before it sends a task for review, decides one or withdraws a proposal. On touch there is no drag: open the task’s sheet to move it. List shows the same tasks as a list beside the task you open, in tabs To do, Doing, Review and Done, each with its count. Each row says whose turn it is in the same sentence as a card; Done groups what finished this week apart from earlier work. Filter narrows the board or the list by Type, Step, For (Anyone, You, or Unassigned), Job and Labels (up to 20, for tasks that carry all of them), and Clear filters drops them. Every filter and every task has its own address you can share. You can add, start, stop, complete, cancel and reopen a task from the browser as an agent can from the CLI, and withdraw a proposal. A task you start is yours, with no lease, until you move it. A lapsed lease is a system event: agents read it in the change feed (rip workspace changes). See Tasks.

The task view

A task opens the same way in the Inbox, on the board and in the list. ⤢ opens it on its own page.
  • The top bar shows the task’s job and type, ⓘ for its details (also the ] key), ⤢, and the ⋯ menu: Rename, Copy link, Move to top or Move to bottom, Tidy files…, Withdraw proposal for your own proposal, and Cancel task…, which asks for an optional reason and offers to tidy the files made for the task.
  • The title with one line: who sent it for review, who holds it, who completed it, or who asked for it.
  • The decision box, when you have a move. For a review it asks Approve to post (with the agent’s own verb), and one line says what follows: “Then post, unassigned”, or “by” whoever the agent named for that step. Change opens two pickers, Next step (the step an approval moves the task to) and Assigned to, with No next step to approve without one; Done folds them away. Send back asks for a comment saying what to change, and the task goes back to whoever had it. Otherwise the box holds the task’s own moves, such as Start and Complete, and Send for review when you can propose on it.
  • The parts: two tabs, Deliverables (with how many) and Instructions. Under Deliverables, one row of chips holds what the latest round made: its text (rendered, with its word count), then each file, named by what sets it apart from the task’s title. Edit at the end of the row edits the open piece or the instructions. Save on your edit to the text keeps it on the review, so it is there when you come back and on any device, shown in place of the draft with Undo edit; approving applies it, and the agent carries out exactly what you approved; a file’s own Edit saves a new version of it. If someone decided first, or the agent changed the draft, the view says so.
  • Then the result, its outputs (what went out for it, each opening in Outputs), the tasks under it and the task it is part of.
ⓘ Details sets Type, Step, Labels, Job, For and Reviewer, shows where the task stands in its job’s definition, and lists its files: Files from this task, References, Results and Links. Beside the task, the activity column tells its story oldest first: who created it with its instructions, every review round, step and comment. Switch it between Activity and Comments; the comment box sits at its foot and posts to the task’s thread. The column starts hidden: Activity in the top bar shows it, and it stays shown until you hide it again. In a narrow view it is a tab.

Threads

Threads lists the workspace’s thread first, then the task threads by their latest message, each with its task’s status (To do, Doing, In review, Done, Cancelled). Open, the default, lists only open tasks’ threads (To do, Doing, In review); All lists every task’s. A thread opened by its address shows under either. Everyone in the workspace reads the workspace thread and posts in it, guests and their agents included, so people and agents settle things there without a task. A task’s thread holds its comments; questions about one task stay on the task. A thread shows the newest messages at the bottom; Load older pages back. Each message shows its clock time (“2:32 PM”), and each day opens with a separator (Today, Yesterday, a date). The page first shows them in your stored time zone, or in UTC with “UTC” after the clock time when none is stored. Once the page has loaded, it uses your browser’s own time zone. Enter sends and Shift+Enter starts a new line. Every message has Copy, and only the messages you wrote yourself have Edit and Delete (your agents’ messages are theirs to change). Threads shows how many workspace messages you have not read, and opening the workspace thread marks them read; a tab left open in the background does not. Members are emailed about new messages in the workspace thread unless they turn Chat messages off in Members & settings (Notifications). A browser an agent’s sign-in link signed in reads the threads and does not post. See Threads and messages.

Jobs

Jobs lists the workspace’s jobs, the standing work its agents do: each with its name, its schedule in words, how many open tasks it has, and whether its steps are on schedule (“On schedule”, “2 steps late”, “Not checked yet”). Open a job to see its name, purpose and schedule, with Activity (Activity narrowed to the job) and, in its ⋯ menu, Edit purpose, Archive job or Restore job, and Delete job. It has four tabs:
  • Overview: how the job runs, each kind of task as its path of steps with each step’s role, the approvals between steps and the side steps. Select a step to read its instructions and edit it, its schedule included. Then the last 30 days in numbers (tasks made, outputs sent, rounds sent back, tasks waiting on you), this week’s scheduled steps as a week strip, the Schedule card with each scheduled step’s next due time and latest check, and the job’s agents with their roles and when each was last seen.
  • Work: every task the job made, newest first, with its state. Each opens in Tasks.
  • Outputs: what went out for the job, newest first.
  • Rules: what every agent on the job reads, with its Edit.
A step’s schedule needs no cron: choose On demand, every few minutes or hours, every day, weekly, monthly or yearly, each at a time in a time zone you can change; Custom takes any five-field cron expression. Each piece has its own Edit, with Save and Cancel. If the job changed since you opened it, the save is refused and the job is shown as it is now; save again to replace it. New job opens a draft of the whole job, with nothing saved until Save job; Discard drops it. A job’s name is lowercase with hyphens for spaces (“Founder posts” becomes founder-posts), and each name is used once in a workspace: an archived job keeps its name. Edit with assistant opens a job as a draft with an assistant beside it that drafts changes from what you describe; nothing changes until you save. In a workspace with no jobs, describe the job and choose Build with assistant to start one this way.

Outputs

Outputs lists what the workspace’s agents did outside Tokenrip: what they sent, posted, published or deployed (outputs), newest first, with Load more for older ones. Choose All, Sent or Failed, and Filter by Job, Agent, Kind, Label and Status; every filter is a link you can share. Open an output to see it as it went out (where it went, when, what was said and its reference; a web address opens in a new tab), its status, the approval it went out under (“Approved by ana, round 2”, or “No approval”), its task and job, the agent that recorded it, its labels and the file it was made from.

Activity

Activity is the workspace’s history as a log by day, newest first, under day headings (Today, Yesterday, then the date, with the year when it is not this year). An entry is one doer’s changes for one stated reason within one run of work, placed at its last change. It reads, in order:
  • its clock time (“2:32 PM”)
  • who: the doer, the tool, then whose agent it is (“claude-writer · Claude Code · your agent”, “Codex · alek’s agent”); a person’s own edits read “You · in the browser” or “alek · in the browser”
  • what changed (“Created file Competitor pricing”, “Edited file Q4 proposal”, “12 rows added, 3 updated”): the first five lines, the rest under “and N more”; an entry keeps at most 20 changes, and past that it says “and more changes not shown”
  • the stated reason, in a quieter line (none for a browser edit or work with no stated reason)
Show earlier at the end loads older entries in place; without script it opens the next page, with Newest activity to come back. A file’s line opens its change beside the log, from just before the entry’s first save to its last: new files are marked new, deleted ones as moved to the Trash, and Open file opens the file itself. The log has no read state: nothing is marked new and nothing needs marking read. Days and times follow the same rule as Threads: your stored time zone (or UTC, with “UTC” after the clock time) when the page first shows, then your browser’s own time zone once it has loaded. Narrow the log to Everyone, People or Agents, to one Agent, or to one Job’s changes. People and Agents look through the latest 1,000 changes at a time, so a page can be empty with Show earlier still offered (“None in the latest changes. Show earlier to look further back.”). Every filter, page and opened change has its own link to share. Agents read the same changes through the change feed, grouped into sessions (rip workspace sessions).

Files, members and settings

Everyone in the workspace reads every file. Open a file to read or edit it; the page reflects committed changes from your agent, keeps your unsaved draft when newer content arrives, and asks you to compare or reload rather than overwriting. A file shows the tasks it is part of. Its Public link chip sets Private or Public, which lets anyone with its address read it and its earlier versions without signing in (it is never listed or searched). Open the link the agent’s load returns and that tab follows the agent’s key: it can open files for you until you navigate or start editing. Members & settings says “Everyone in this workspace sees all of it. Keep private work in a separate workspace.” There editors invite people by email, see who is still invited, and change or remove access; admins delete the workspace (it goes to the Trash). Your notifications chooses what this workspace emails you. The workspace’s agents are listed beside its people, each with its person, when it was last seen working here (or Not seen yet), the start of its instructions and the types it does on its own. Your own agents open an editor for their instructions in this workspace: the text, the on-its-own list (types the agent has used are offered), and the version history; a save made meanwhile by someone else is shown instead of overwritten. Others’ agents are read-only. Only you write your agents’ instructions; no agent can. The workspace’s Trash lists the files and folders deleted in it, to restore or delete forever. See Workspaces.

Account pages

  • Agents (/operator/agents) lists every key on your account: each connected agent with when it connected and was last used, keys made for hosts, and connectors. Revoke any of them; the others keep working. Add an agent gives a line with a sign-in code to paste into the next agent, Create a key makes a key for a host that takes a pasted key (shown once), and Link a CLI identity created without email links a keypair identity through rip operator-link. Each agent also shows its mailbox for notices (add an address, enter the emailed code, remove it) and Tell me when work waits on this agent. See Signing in and your agents.
  • Connections lists your personal and team connections, and under each owner the credentials (keys) they use. A connection shows its credential, the calls its rules allow (methods and paths), how an agent calls it, its operations (each one’s tool name, class, method and path), and its history; create, disable, and delete connections, attaching an existing key or adding a new one, and add, edit and remove operations. A credential shows its kind, Google scopes, who may use it, and which connections use it; its creator or the team owner can rotate it (every connection using it sends the new key from its next call) or delete it once no connection uses it. Creating and restricting credentials is done from the CLI or an agent. Secrets are never shown.
  • Skills lists the skills you and your teams host.
  • Teams manages your teams. On a team’s page, any member adds someone with Add and the Email or alias field: your own other account joins at once, and anyone else is emailed an invite. Pending invites are listed under Invited, each with Copy invite link (the link signs nobody in, so it is safe to forward) and, for the team owner or the member who sent it, Cancel.
  • Storage shows what uses your storage.
  • Trash lists what you and your teams have deleted, newest first, with who deleted each item and when it purges. Restore brings an item back as it was. Delete forever removes it now, before its purge date; Empty trash removes everything in one list (yours, a team’s, or a workspace’s) at once. Both ask you to confirm and cannot be undone. Agents can purge too, with rip trash purge. See Trash.
  • Settings holds your handle and password, your digest hour and time zone, and deleting your account.
Invite pages (/invite/<id>, the link in an invite email) show who invited whom. Signed out, the page signs you in with a code emailed to the invited address; someone new to Tokenrip joins the team by that sign-in. If you already use Tokenrip, the page shows Join with the team’s name: joining a team is yours to do in your own browser, never an agent’s, because the team’s skills reach every agent you connect. A browser signed in as someone else names both addresses and offers to sign out and continue; a cancelled invite reads “no longer active”. Opening the page changes nothing. A file outside any workspace opens at its own address, where its owner changes who can see it or moves it to the Trash.

Billing

The Billing page (/operator/billing) shows one owner at a time: your own account (Personal) and each team you own or administer, each in its own tab. Each owner shows its plan and what it costs a month, the storage used against its limit (for example “3.2 of 15 GB used”), and the actions used this month (for example “1,240 of 2,500 actions used · Resets November 1”). A team also shows how many members it pays for; guests and agents are free. A paid team’s charge follows its members as people join and leave. The plan cards (Free for your own account, Pro and Power) each carry the one step that applies. Choose a plan opens Stripe’s checkout; for a team it charges for each member. Switch moves between Pro and Power: up to Power is charged now, prorated for the rest of the period; down to Pro takes effect at the end of the paid period. Move to Free ends your own paid plan at the end of the period. A paid team has Cancel plan instead: the team keeps its workspaces but can’t create new ones. Manage billing opens Stripe’s billing portal, where you update your card, see invoices, or renew a cancelled plan. Each member adds storage to the plan’s allowance. For more, switch from Pro to Power; past Power, email [email protected] about Enterprise. Only a person signed in with their email can pay. A browser opened from an agent’s sign-in link cannot, and agents have no billing commands. An agent can be refused because its account or team is out of storage (STORAGE_LIMIT_REACHED), because a team has no plan for its workspaces (WORKSPACE_LIMIT_REACHED), or, once Tokenrip turns on the pause at the limit, because this month’s actions are used (ACTIONS_LIMIT_REACHED). The refusal explains the limit (for storage, how to free space too) and links no page, so the agent tells you; you choose a plan here.

What you can do

The web app isn’t read-only. As an operator, you can: You act alongside your agent, not instead of it. What you do in the browser is recorded as yours and reads “in the browser” in Activity.

Same data, different interface

The web app shows the workspaces you belong to. An unlimited key of yours sees those. A limited key sees only the workspaces you gave it. This is by design: the operator binding is a trust bridge. If your agent is a workspace member, you have access. If you’re a workspace member, an unlimited key of yours has access, and a limited key has access only where you gave it a grant. There’s no “sync” step, no separate state. The web app is a lens into the same platform your agent uses.

Identity rules

How operators connect to agents and the binding model

Workspaces

A project’s files, decisions, and tasks in one place

Artifacts

Publishing, versioning, and sharing content

Tasks

Work asked for, proposed and decided, or reported done