ucitra/ agent teams

Commands

Run lucitra with no arguments to open the shell, or lucitra <command> to run one and exit. The same commands are served to other agents by lucitra mcp, and every one of them is checked the same way wherever it comes from.

Reads

Change nothing. Safe from any surface, including a read-only bridge.

CommandArgumentsWhat it does
plangoalScope a goal into tasks without running any of them
chatmessageAsk your chief of staff something, in this repo’s conversation
askquestionSearch the project brain, falling back to the chief of staff
surveyWhat Agent Teams can work out about this repo without being told
candidatesWork this repo already says needs doing: issues, drafts, old TODOs
routegoalAsk the chief of staff who should take a request. Starts nothing
reposThe repositories a request may start in, this session’s first
tasksWhat has run and what is running, newest first
dashboardWhere the work stands across every repository, and what the week cost
threadswhereEvery thread in this repo: what ran, what needs you. everywhere for all
searchtextFind a line across every conversation in this repo
modelsWhich model each agent runs, and on which runtime
mcpWhich MCP servers each agent may reach
agentsThe roster, and what each agent is configured to do
companyThe team this repository declares, and who reports to whom
org-chartEvery org chart on this install, and which one delegation is checked against
validate-teampathRead an agent team from a folder and list what is wrong with it
integration-healthrefreshWhether each integration and key source works, and the fix when not
schedulesEvery schedule the agent teams declare, whether it is on, and what it last did
brain-loose-linksLinks from the brain files that are not pages: the index, the log
trigger-previewscheduleWhat each trigger would fire on now, without firing: [<team>/<project>/<task>]
identitycompanyWho a company’s chief of staff is here: its name, photo and look
identitiesEvery agent’s name and photo on this install
budgetsEvery token budget, and what each team or agent has spent this period
trading-limitsTrading limits every order is held to: defaults, yours, each team’s
secret-sourcesWhere integration secrets are read from: GCP projects and env files, in order
model-providersWhere models can come from besides local: your own keys for a provider
run-eventsrunOne run, step by step: every tool call, its output and what the agent said
connectorsrefreshMCP servers your runtimes connect to themselves: claude.ai and config
integrationsWhich integrations the agents name, and whether Agent Teams has them
brain-pagesEvery page in the brain: its frontmatter and what it links to
brain-pagepathOne page of the brain, text and all: brain-page <path>
brain-historypathEvery commit that touched the brain: brain-history [path] for one page
approvalsWhether runs ask before anything leaves this machine, or act on their own
brainsBrains registered on this machine, and which one brain reads use
browser-hostWhether the Companion’s native messaging host is registered here
first-runThe queues and the places first run offers
preview-companyqueue, placeThe team a queue and a place would produce: preview-company <queue> <place>
channelsWhere Agent Teams may reach you when you are not at the terminal
accountWho this install belongs to, and what proves it

Writes

Change something on this machine. A bridge refuses these unless it was started with --allow-writes.

CommandArgumentsWhat it does
seedkeepWrite what the survey found into this repo’s brain, as one page
start-buildgoal, agent, repo, team, note, pushStart a task in its own worktree, as the Builder or a named agent
replythread, messageReply in a thread; the agent that last ran there picks it up
declinetaskRefuse a waiting push. The worktree is left alone
teamsteamThe agent teams on this install; with a slug, switch to one
scheduleschedule, onOffTurn one schedule on or off: schedule <team>/<project>/<task> on|off
cancel (destructive)taskStop a running task’s agent. Its worktree is left as it is
edit-agentagent, field, valueChange a team agent’s name, title, skills or instructions
set-identityagent, field, valueGive an agent a nickname or photo: <team> for its chief, or <team>/<role>
clear-identityagent, fieldPut an agent back to its role name, or its default look
set-company-namenameRename your default company. A team takes its name from its package
run-schedulescheduleRun one schedule now, whether or not it is on: <team>/<project>/<task>
budgetteam, tokens, period, agentSet or clear a token budget: <team> <tokens|off> [day|week|month] [agent]
integration-addintegration, environmentInstall an integration, so agents granted its tools can call them
trading-limitintegration, key, valueChange a trading limit: <alpaca> <key> <value|default|none>
trading-limit-tightenintegration, key, valueTighten a trading limit; refuses a change that loosens it
secret-source-add`keychain1password
model-provider-setprovider, setting, valueSet one model provider setting: <provider> <SETTING> <value|none>
secret-source-remove`keychain1password
integration-removeintegrationRemove an integration; its tools leave every agent from the next run
approval-mode`askauto`
brain-addpath, nameRegister a folder kept as a brain: brain-add <path> [name]
brain-usenameRead from a registered brain, or repo for this repository’s own
brain-removenameForget a registered brain. Its folder is left as it is
declare-companyqueue, placeWrite the team a queue and a place would produce to .lucitra/company.md
model-preset`claudewithout-claude`
configureagent, setting, valueChange one setting on one agent: configure <agent> <setting> <value>
resetagent, settingPut one setting back to its default: reset <agent> <setting>
runtimesruntimeWhat runtimes are installed; with an argument, switch to one
verify-channelchannel, codeProve a channel by echoing back the code it was sent
revoke-channel (destructive)channelStop sending to a channel. That it existed is kept
verify-claimclaim, codeEcho back the code an identity was sent
revoke-claim (destructive)claimStop an identity proving this account. That it did is kept
new-conversationtitleBegin a new conversation, optionally named
conversationsconversationList conversations; with a number, reopen that one

Leaves the machine

Sends something outward: a push, a message, a verification code. Never run unattended without meaning to.

CommandArgumentsWhat it does
approvetask, approvedLet a waiting task push. Omit the task when only one is waiting
team-addsourceRegister an agent team from a folder, or clone one from github:owner/repo
team-updateteamPull the latest of a team's folder, fast-forward only, and refresh its roster
add-channelkind, address, labelAdd somewhere to reach you: add-channel <ntfy|sms|whatsapp> [address] [label]
claimkind, valueProve an identity for this account: claim <phone|email> <value>

83 commands: 42 read, 35 write, 6 leave the machine.

Cookie settings

Choose what this site may store in your browser. You can change it any time from Cookie settings at the bottom of every page.

EssentialAlways on

Remembers the choice you make here. Nothing else is stored until you turn analytics on.

What is stored
NameWherePurposeKept
lucitra-consentLocal storageYour analytics choiceUntil you clear it
Analytics

Shows us which pages people read, what breaks, and which links bring people here, so we know what to fix. Off until you turn it on.

What is stored
NameWherePurposeKept
ph_<key>_posthogCookieA random identifier for your browser, and the current session1 year
ph_<key>_posthogLocal storageThe same identifier, as a backup to the cookieUntil you clear it
lucitra-last-touchSession storageThe link or campaign that started this visit, sent only with a contact form messageUntil you close the tab
lucitra-first-touchLocal storageThe link or campaign of your first visitUntil you clear it
lucitra-last-touchLocal storageThe link or campaign of your latest visitUntil you clear it
Cookie policy