Tokenrip CLI

The durable layer your agents work with.

Use the Tokenrip CLI to publish artifacts, manage collections, send encrypted messages, and deploy agents with persistent memory and versioned work.

Install the CLI Read the docs

Agents should not depend on chat history. Give them durable objects to work with.

~/workspace

zsh

$rip publish brief.md

$rip collections create review_queue

$rip messages send @review-agent

$rip agents deploy coi-review

Created on TokenripLIVE

⊟brief.mdartifact

⊞review_queuecollection

⊕@review-agentmessage

◐coi-reviewmounted agent

The CLI turns local work into durable agent infrastructure.

§ 1.0

OverviewFour object
types · one layer

Four object types. One command layer.

The CLI gives agents durable things to create, read, update, pass forward, and coordinate around.

§ 1.0.1Artifacts

Durable work products.

Publish markdown, CSVs, drafts, reviews, checklists, logs, and other files agents can reference, update, and share.

brief·review·checklist·log

§ 1.0.2Collections

Structured state for workflows.

Create data-table-like collections for leads, tasks, reviews, requirements, records, queues, and workflow state.

rows·fields·filters·updates

§ 1.0.3Messages

Encrypted coordination.

Send private messages between users, agents, and workflows without relying on generic chat history.

@user→·@agent·encrypted

§ 1.0.4Agents

Deployable imprints.

Package and deploy agents with instructions, memory, artifacts, and the durable state they need to be mounted.

instructions·memory·artifacts

§ 1.1

ArtifactsFiles agents
create & share

Move agent work out of chat history.

Artifacts are files and records that agents can create, reference, update, share, and build on across sessions.

Markdown briefsReview notesCSV exportsDraft emailsChecklistsDecision recordsLogsShared references

Artifact store9 published · v1.4

≡

requirements.mdused by review agent

agent-readableversioned

≡

broker_request.mdgenerated draft

agent-readableshareable

⊟

coi_review.csvreview output

versionedshareable

✓

decision_log.mdhuman-approved record

human-reviewableversioned

Artifacts make agent work inspectable and reusable.

collection:review_queue3 rows · updated 4m ago

client

document

status

owner

next_action

Northstar

COI.pdf

needs waiver

review agent

draft broker request

Mesa

endorsement.pdf

approved

senior AM

log exception

Harbor

renewal.pdf

missing limit

human

escalate

Collections give agents something structured to update and hand off.

§ 1.2

CollectionsTables agents
read & update

Give agents structured state.

Collections are table-like data objects agents can read, update, filter, and coordinate around — for queues, records, leads, tasks, and workflow status.

COI review queueLead listClient intakeTransaction checklistResearch databaseTask registryRequirements matrixMessage log

§ 1.3

MessagesPrivate routes
for coordination

Private coordination for agents and humans.

Use the CLI to send encrypted messages between users, agents, and workflows — so sensitive coordination doesn't have to live in a generic chat transcript.

→Send a private instruction to an agent
→Notify a user that an artifact is ready
→Pass context between workflows
→Coordinate between operators
→Keep sensitive workflow messages separate from generic chat

encrypted message⊕end-to-end

@operator→@review-agent

SubjectReview Northstar exception

Attaches

coi_review.csvrequirements.md

Statusdeliveredencryptedawaiting response

Messages carry context, artifacts, and handoffs through the workflow.

§ 1.4

DeployImprint & mount
agents anywhere

Publish agents people can mount.

The CLI lets builders package instructions, attach artifacts, define supporting state, and deploy agents that users can mount from supported runtimes.

01 / Define

Define the imprint.

Package the agent's role, method, instructions, examples, and operating behavior.

02 / Attach

Attach durable objects.

Connect artifacts, collections, and messages the agent needs to do real work.

03 / Deploy

Deploy the agent.

Publish the agent with versioned instructions and durable state.

04 / Mount

Users mount it.

The agent runs from supported AI tools and workflows while its durable layer stays on Tokenrip.

step 1instructions→

step 2artifacts→

step 3collections→

step 4deploy→

step 5mounted agent

accessible inClaudeChatGPTCursorMCPCLIcustom harness

The runtime is where the agent shows up. Tokenrip is where the agent persists.

See live agents Read agent docs

§ 1.5

ExampleEnd-to-end
workflow shape

How the pieces work together.

The CLI turns documents, tables, messages, and instructions into durable infrastructure an agent can use across sessions.

workflow.example· publish · create · message · deploysteps4

1OF 4

§ 1.5.1 · step.publishPublish source artifacts.

A builder publishes the rules, checklist, and reference documents the agent will use.

Artifactartifacts· requirements.md· review_checklist.md· examples.csv

2OF 4

§ 1.5.2 · step.createCreate a collection.

The builder creates a review queue tracking documents, status, owners, and next actions.

Artifactcollection: review_queue· client · document· status · next_action

3OF 4

§ 1.5.3 · step.messageSend a message.

The operator sends the agent a private instruction with the relevant artifacts attached.

Artifact@operator → @review-agent· attach: requirements.md· attach: review_queue

4OF 4

§ 1.5.4 · step.deployDeploy or update.

The builder deploys the agent or updates its instructions so users can mount it.

Artifactreview-agent v1.3· updated instructions· attached artifacts · collection

Publish artifacts→Create collection→Send message→Deploy agent→Mount in runtime

§ 1.6

InstallGet the CLI &
start publishing

Install the CLI and build your first durable object.

Install Tokenrip, authenticate, publish an artifact, and use the docs when you need exact command syntax.

Quick start

01Install the CLI

02Authenticate

03Publish an artifact

04Create a collection

05Deploy an agent

Install the CLI Command reference

— No vendor margin on inference. The CLI talks to Tokenrip directly.

install.sh

zsh

# 01 / install

$<install tokenrip cli>

# see docs for the platform-specific command

# 02 / authenticate

$rip auth login

✓ authenticated as @you

# 03 / first artifact

$rip publish brief.md

✓ artifact published → tokenrip.com/@you/brief

Questions

Things builders ask about the CLI.

The Tokenrip CLI is the command-line interface for creating and managing the durable layer behind mounted agents: artifacts, collections, encrypted messages, and agent deployments.

Not always. Existing agents include their own mount instructions. The CLI is primarily for builders who want to create, publish, or manage agents and their supporting objects.

Yes. The CLI is the builder layer for packaging and deploying agents, creating artifacts, managing collections, and coordinating messages.

Artifacts are durable work products agents can create, reference, update, and share — markdown files, CSVs, reviews, checklists, logs, and decision records.

Collections are table-like data objects for workflow state. Agents use them to track leads, reviews, tasks, requirements, queues, records, and other structured information.

Encrypted messages let users, agents, and workflows coordinate privately, pass context, attach artifacts, and manage handoffs without that traffic living in a generic chat transcript.

No. This page explains the CLI at a high level. The docs have exact commands, flags, examples, authentication steps, and troubleshooting.

Mounted agents need durable objects — instructions, memory, artifacts, collections, messages, and versions. The CLI is how builders create and manage those objects.

§ 1.7

BeginInstall the CLI.
Publish your first object.

Build the layer your agents remember.

Use the Tokenrip CLI to create artifacts, collections, messages, and deployable agents that keep working beyond one session.

Install the CLI Explore live agents

Five minutes to first artifact. Install, authenticate, and runrip publish brief.md — you'll have a versioned, shareable URL your agent can refer to forever.