Why origami-cli

• The whole API. All 120 v3 operations are commands, built from Origami's OpenAPI spec. When Origami ships a change, pnpm sync:v3 regenerates the catalog; nothing is hand-maintained.
• Checks before it sends. Enum values and required fields are validated before any request leaves your machine. launch --dry-run shows the exact blockers a real launch would hit.
• Safe to retry. Every POST carries an Idempotency-Key that is reused across automatic retries, so a flaky network never launches a campaign or enrolls a list twice.
• Built for scripts. JSON when piped, tables in a terminal, CSV on request. --all follows cursors to the end, and data goes to stdout while progress goes to stderr.
• Async done right. Anything that returns a Job takes --wait, which polls on the Job's own next_poll_at. origami jobs wait <id> picks up any Job later.

Install

npm install -g @braingrid/origami-cli     # puts `origami` on your PATH
npx @braingrid/origami-cli --help         # or run it without installing

Requires Node.js 18+. From source: pnpm install && pnpm build && npm link.

Quickstart

origami auth login                                          # paste an og_live_… key
origami send campaigns list --status draft                  # what's waiting to go out
origami send campaigns examples list <campaignId>           # the copy each person will get
origami send campaigns launch <campaignId> --dry-run        # blockers + held-back counts
origami send campaigns launch <campaignId>                  # go
origami send campaigns stats <campaignId>

Authenticate

Create an API key under Settings → Developers in the Origami app. Keys look like og_live_… and are parent-wide (they can act on your org or any of its projects).

origami auth login                    # prompts for a key (hidden input), verifies it
origami auth login --key og_live_…    # non-interactive
echo "$KEY" | origami auth login      # from a pipe / secret manager
origami auth login --project <id>     # store a default project (child org) scope

Inspect or rotate:

origami auth status                   # resolved profile (key masked) + plan
origami auth logout                   # forget the active profile's key

The key can also come from the environment (handy for CI):

export ORIGAMI_API_KEY=og_live_…
origami account

The core loop

Origami is agentic: you send a brief, an agent researches it in the background and builds a table, then you read the rows.

# 1. Run a brief (creates an agent, polls to completion, prints the table)
origami run "Find 20 fintech CEOs in NYC" --model max

# 2. If the agent asks a question, answer it (same agent + conversation)
origami agents ask <agentId> "Only Series A and later, please"

# 3. Read the rows it built
origami tables rows <tableId>
origami tables rows <tableId> -o csv --out leads.csv     # export to a file

# 4. Bring your own data — upsert + enrich
origami tables upsert <tableId> --match domain \
  --row '{"domain":"acme.com"}' --row '{"domain":"globex.com"}'
origami enrichments get <enrichmentRunId> --wait          # poll until settled

origami run and origami agents create are the same flow; run is the friendly shortcut. Add --no-wait to return the running run id instead of polling, then poll later with origami agents run-get <agentId> <runId> --wait.

Global options

Place these after the command (kubectl-style):

Data goes to stdout; status, progress, and hints go to stderr — so | jq and -o csv pipelines stay clean.

Commands

run / agents

origami run "<brief>" [--model lite|max] [--workspace <id>] [--table <id>]... [--rows] [--no-wait]
origami agents list [--search <text>] [--all]
origami agents get <agentId>
origami agents ask <agentId> "<follow-up>"
origami agents runs <agentId>
origami agents run-get <agentId> <runId> [--wait] [--include stats,transcript]
origami agents tables <agentId>
origami agents cancel <agentId>
origami agents archive <agentId>

tables

origami tables list [--workspace <id>] [--all]
origami tables get <tableId> [--include stats]
origami tables columns <tableId>
origami tables rows <tableId> [--filter '<col> <op> [value]']... [--sort <col>:<asc|desc>] \
                              [--flat] [--no-defaults] [--limit <=200] [--all] [-o csv] [--out file.csv]
origami tables row <tableId> <rowId>
origami tables cell <tableId> <rowId> <columnId>
origami tables upsert <tableId> --match <slugs> (--row '<json>'... | --file rows.csv) \
                              [--no-enrich] [--reenrich-updated] [--batch-id <uuid>]

Filter operators: contains, not_contains, equals, not_equals, is_empty, is_not_empty, greater_than, greater_than_or_equal, less_than, less_than_or_equal.

enrichments, documents, workspaces

origami enrichments list [--table <id>] [--all]
origami enrichments get <runId> [--wait]

origami documents list <workspaceId>
origami documents upload <workspaceId> <file...> [--mode table|append|document] [--table <id>]
origami documents get <workspaceId> <documentId> [--out <file>]
origami documents rename <workspaceId> <documentId> <name>
origami documents delete <workspaceId> <documentId>

origami workspaces list
origami workspaces create --name <name>
origami workspaces get <workspaceId>
origami workspaces delete <workspaceId> [--confirm]

v3: send, leads, jobs, account

Each operation id is the command path: send.campaigns.people.list is origami send campaigns people list, fetch_more becomes fetch-more. A group whose only operation is get collapses into the group (origami send campaigns stats <id>), and an argument-free get is its group's default (origami account, origami account credits).

# The send recipe, in order
origami send campaigns create --name "Q4 delivery leaders" --channels linkedin
origami send campaigns people upsert <campaignId> --list-id <listId>
origami send campaigns templates put <campaignId> --data @template.json
origami send campaigns settings patch <campaignId> --no-auto-lead-refill-enabled --require-message-approval
origami send campaigns senders add <campaignId> --sender-id <senderId>
origami send campaigns examples list <campaignId>                # read the rendered copy per person
origami send campaigns launch <campaignId> --dry-run             # blockers + held-back counts
origami send campaigns launch <campaignId>
origami send campaigns stats <campaignId>

# Leads
origami leads lists list
origami leads lists rows list <listId> --all -o csv > rows.csv
origami leads lists rows upsert <listId> --rows @rows.json --match-columns email
origami leads searches create --brief "…" --count 25 --wait

# Jobs: every 202 returns one; --wait on the command, or poll later
origami jobs list --status running
origami jobs wait <jobId>
origami jobs input <jobId> --answers "Only Series A and later"

# Account
origami account senders list --channel linkedin
origami account rate-limits
origami account exclusion-lists people add --email someone@example.com

How flags map to the API:

• Path parameters are positional arguments; query parameters and top-level body fields are --kebab-case flags. origami <command> --help lists them with the spec's descriptions and ends with the HTTP call and its docs page.
• Booleans take --flag or --no-flag (sends false). Lists take commas or repeats (--ids a,b --ids c). Object fields take JSON inline, @file.json, or - for stdin.
• --data <json|@file|-> supplies the whole body; flags override its keys. A body field that would shadow a global option gets a body- prefix (--body-fields).
• Enums and required fields are checked before any request is sent.
• Every POST carries an Idempotency-Key (yours via --idempotency-key, else a UUID) that is reused across automatic retries, so a retried launch or enrollment never runs twice.
• List commands print one page and hint the cursor; --all follows next_cursor to the end.
• Operations that return a Job accept --wait (polls on the Job's next_poll_at, default ceiling 30 minutes, --wait-timeout <s> to change it).

Regenerate the command catalog when Origami ships spec changes:

pnpm sync:v3              # fetches https://docs.origami.chat/openapi-v3.yaml → src/v3/catalog.ts
pnpm sync:v3 ./spec.yaml  # or from a local copy

campaigns, sequences (v2)

origami campaigns list (--workspace <id> | --table <id>)
origami campaigns create <tableId> "<instructions>"     # agentic — drafts sequences, doesn't send
origami campaigns edit <campaignId> "<change>"          # agentic
origami campaigns people <campaignId> [--status <csv>] [--all]
origami campaigns stats <campaignId>
origami campaigns launch <campaignId> [--dry-run]       # also: pause / resume
origami campaigns delete <campaignId> [--confirm]

origami sequences list (--workspace <id> | --table <id> | --column <id>) [--status <s>]
origami sequences get <sequenceId>
origami sequences stop <sequenceId> [--dry-run]
origami sequences delete <sequenceId> [--force]

scheduled, projects (v2)

origami scheduled list [--workspace <id>] [--enabled true|false]
origami scheduled create --workspace <id> --name <n> --prompt "<brief>" --cron "0 9 * * 1"
origami scheduled enable|disable|trigger <id>
origami scheduled runs <id>

origami projects list                                   # parent-org only
origami projects create <name> [--monthly-credits <n>]
origami projects update <id> [--name <n>] [--monthly-credits <n|null>]
origami projects delete <id> [--confirm]

origami credits                                         # credit balance (v3 shortcut)

webhooks (local dev tools)

Origami webhooks are configured in the dashboard; these are local helpers for building a receiver.

origami webhooks verify --secret whsec_… --id <id> --timestamp <ts> --signature <sig> --file body.json
origami webhooks listen --port 9333 --secret whsec_…    # verify + pretty-print inbound events

config

origami config path
origami config show                                     # secrets redacted
origami config profiles
origami config use <profile>

Scripting

origami tables rows <id> -o json | jq '.[].website'
origami agents list --all -o csv > agents.csv
origami run "Find 50 founders" --rows -o json > result.json

Reliability

The client automatically retries 408/429/500/502/503/504 and network errors with exponential backoff (honoring Retry-After), and polls runs on the API's Retry-After cadence (15s). Rate limits: 300 req/min per IP, 100 req/min per org; concurrent agent runs are plan-tunable.

Exit codes

Environment variables

Development

pnpm dev -- <args>        # run from source (tsx)
pnpm typecheck            # tsc --noEmit
pnpm test                 # vitest
pnpm build                # tsup → dist/index.js
pnpm sync:v3              # regenerate src/v3/catalog.ts from the published v3 spec

Contributing

Issues and pull requests are welcome. CONTRIBUTING.md covers setup, the test suite, and how the v3 catalog is regenerated. Please report security issues privately as described in SECURITY.md.

Built by the team behind

   

_{BrainGrid: the app builder that plans before it builds · Plansmith: the planning agent for the business side of software · Radial: the issue tracker for coding agents}

License

MIT. origami-cli is an independent project and is not affiliated with or endorsed by Origami. "Origami" is a trademark of its owner.