Command-line interface for Index Network. Message your personal agent, manage intents, and discover opportunities — all from your terminal.

npm install -g @indexnetwork/cli

Index helps you find the right people—and helps the right people find you—based on what you are actually trying to do, not just a profile headline. The value is grounded intros: suggestions come from communities you share (syndicates, founder groups, firm networks), not from spraying the open web.

The flow below is one complete story—create an intent, let agents negotiate in the background, then review a persisted match.

# login + setup
index login
index profile

# 1. express an intent
index intent create "federated learning collaboration"

# 2. agents negotiate on your intents in the background
index negotiation list

# 3. wait for persisted opportunities, then review outcomes
index opportunity list --status pending
index opportunity show <opportunity-id>
index opportunity start-chat <opportunity-id>

Words you will see elsewhere in this doc: network = a community you are in; intent = your “what I am looking for” post; opportunity = a suggested introduction between you and someone else.

Every exposed CLI command is listed below. Add --json for machine-readable results; conversation stream --json emits one event per line. Command-specific flags are listed in Options, and examples follow this reference.

Authenticate with Index Network. Opens a browser window that runs the device authorization grant against your existing session (or a fresh login), then hands this machine a session of its own.

index login local               # http://localhost:3001 API + http://localhost:3000 web app
index login dev                 # https://protocol.dev.index.network API + https://dev.index.network web app
index login prod                # Production (also the default on first login)
index login --api-url <url>     # Custom API; matching web origin is inferred
index login --api-url <url> --app-url <url>  # Override both origins

Credentials are stored in ~/.index/credentials.json; the latest login replaces the previous target and its session. Without a target, login reuses the stored API origin (or production on first use). The local web app must be running on port 3000 and the API on 3001; the dev web app must be configured to authenticate against the dev protocol service. Browser login explicitly requests protocol v2 and binds the loopback callback with a one-time state. Only a short-lived device code travels through the redirect; the CLI exchanges it at /api/auth/device/token for its own session token and sends that as Authorization: Bearer. There is no approval prompt, because the web page mints and approves the code itself — no code from anywhere else can enter the grant. A re-login revokes the session it replaces, so logins do not pile up.

For noninteractive calls, set exactly one credential:

export INDEX_API_URL='https://protocol.index.network'
export INDEX_API_KEY='<api-key>'
index --api-url "$INDEX_API_URL" intent list --json

INDEX_SESSION_TOKEN uses Bearer authentication; INDEX_API_KEY uses x-api-key. Setting both fails. Environment credentials take precedence over stored browser login. The API origin resolves from --api-url, then INDEX_API_URL, then the stored credential URL, then https://protocol.index.network. Integrations pass the origin explicitly. Agent management and other session-only operations still require a session.

index docs                       # Summary plus the list of topics
index docs workflows --json
index agent me --json

Topics are identity-context, intents, communities-networks, opportunities, negotiations, and workflows. An unknown topic exits nonzero.

Revoke this machine's session server-side, then clear the local credential file. A session can revoke itself, so sign-out takes effect immediately without needing your browser. If the server cannot be reached the local file is still cleared, and logout tells you to revoke the device in Index web settings. If local cleanup fails, logout exits nonzero and asks you to remove the file manually.

index logout

Manage your intents. Create intents from natural language, list active intents, view details, and archive intents you no longer need.

index intent list                           # List active intents
index intent list --archived                # Include archived intents
index intent list --limit 5                 # Limit to 5 results
index intent list --query "CTO"             # Match description and summary text
index intent show <id>                      # Show full intent details
index intent prepare "Looking for a CTO"  # Read feedback / recovery prompts and receipt
index intent prepare "Looking for a CTO" --answer 'Prompt label=My answer'
index intent create "<returned payload>" --receipt '<returned preparationReceipt>'
index intent pause <id>                   # Hold the agent without archiving
index intent resume <id>                  # Start it again
index intent update <id> "revised text"     # Update an intent (runs full pipeline)
index intent archive <id>                   # Archive an intent
index intent networks <id>                  # List the networks an intent is shared in
index intent add-to-network <id> <network-id>      # Add an intent to a network
index intent remove-from-network <id> <network-id> # Remove an intent from a network

Preparation returns needs_revision with recovery field labels or ready with a payload and preparationReceipt. Use the exact returned payload with its receipt in create; --answer pairs the displayed recovery label with your response. You can also create directly: the server prepares the description and rejects an unadmitted draft. A new intent is shared in every network you belong to. Narrow it afterwards with remove-from-network, or widen an existing one with add-to-network.

Read current negotiations and submit structured turns using full opportunity IDs.

index negotiation list [--intent-id <id>] [--state open|settled]
index negotiation show <opportunity-id> --json
index negotiation turn <opportunity-id> --action propose --message "..." --json

Actions are propose, counter, accept, and decline. Read the detail's protocol.availableActions and current turnCount before submitting. The server enforces eligibility and stale-count protection. The CLI preserves all turns and protocol guidance and never retries a rejected or uncertain write. Agreement between agents remains separate from the owner's opportunity approval. --limit, --since, and --status are unsupported negotiation filters.

Browse and manage discovered opportunities.

index opportunity list                     # List all opportunities
index opportunity list --status pending    # Filter by one status
index opportunity list --intent-id <id> --statuses negotiating,pending,accepted,expired
index opportunity list --limit 5           # Limit results
index opportunity show <id>                # Show full details
index opportunity accept <id>              # Accept an opportunity
index opportunity reject <id>              # Reject an opportunity
index opportunity start-chat <id>          # Open its human chat (pending/accepted)

--status accepts one of pending, accepted, rejected, expired; --statuses accepts a comma-separated radar selection including negotiating. The intent-scoped radar uses --intent-id <id>.

Manage networks (communities). Network creation is direct for eligible staff; otherwise the command submits an early-access request. Invitations accept any valid email and use the server invitation flow.

index network list                     # List your networks
index network discover                 # Find public networks
index network requests                 # View your creation requests
index network request-update <id> --title "New Name" --prompt "Purpose"
index network request-dismiss <id>
index network create "My Network"  # Create directly when eligible; otherwise submit an early-access request
index network create "AI" --prompt "AI researchers"  # Create or request with a description
index network show <id>                # Show details and members
index network update <id> --title "New Name"  # Update a network
index network delete <id>              # Delete a network
index network join <id>                # Join a public network
index network leave <id>               # Leave a network
index network invite <id> user@email # Invite directly by email

Read and message the personal agent within an intent, or use human DMs.

index conversation show agent --intent-id <id> --json
index conversation send agent "..." --intent-id <id> --json
index conversation send agent "..." --intent-id <id> --question-id <pending-question-id> --json
index conversation answer agent --intent-id <id> --answer 'question-id=answer' --answer 'other-id=answer' --json
index conversation list
index conversation with <user-id>
index conversation show <conversation-id> [--limit <n>]
index conversation send <conversation-id> "..."
index conversation stream --json

Agent reads include availability and the pending question. An answer must carry the current question ID and its intent scope; stale refusals are surfaced directly. Human messages retain their existing conversation IDs. SSE emits one JSON event record per line. There is no local agent runtime or interactive chat session.

index onboarding confirm-profile --json
index onboarding complete [--intent-id <id>] --json

Confirm a profile only after the owner reviews it. Completion uses the server's profile-confirmation and first-intent prerequisites; it does not bypass them. For a new account: profile sync to see suggested facts, profile update to save what you approve, onboarding confirm-profile, intent prepare and intent create with the returned payload and receipt, then onboarding complete --intent-id <id>.

View user profiles and synchronously enrich your public identity, social, and avatar data.

index profile                       # Show your own profile
index profile show <user-id>        # Show another user's profile
index profile sync                  # Run public profile research prefill (does not persist)
index profile update --name "Name" --intro "Bio" --location "City"
index profile update --social 'website=https://example.com' --social 'github=https://github.com/me'

--social replaces the full saved social-link list; include every link you want to keep. Profile confirmation and onboarding completion remain explicit commands.

Scrape and extract content from a URL.

index scrape https://example.com                    # Scrape a URL
index scrape https://example.com --objective "..."   # Scrape with focus

Sync profile, networks, and intents to a local file.

index sync                             # Sync to ~/.index/context.json
index sync --json                      # Output to stdout as JSON

--json emits one parseable result or error on stdout. Progress goes to stderr; HTTP failures exit nonzero. Sync fails if any required context read fails, without saving partial context.

Approved intents are evaluated in the background. opportunity list only reviews persisted results; it does not start evaluation.

# Inspect agent negotiations
index negotiation list

# List persisted pending opportunities
index opportunity list --status pending

# See full details
index opportunity show <id>

# Accept or reject
index opportunity accept <id>
index opportunity reject <id>

# Run directly with Bun (no build step)
bun src/main.ts intent list --json

# Build for all platforms
bun run build

# Build for current platform only (fast dev builds)
bun scripts/build.ts --current

# Dry-run publish
bun scripts/publish.ts --dry-run