Skip to main content

tinyfish search query

Run a web search. Returns ranked results with titles, URLs, and snippets.

Flags

FlagDescription
--location <value>Location hint for geo-targeted results (e.g. "Vietnam")
--language <value>Language hint (e.g. "en")
--prettyHuman-readable output

Output

Examples


tinyfish fetch content get

Fetch clean, extracted content from one or more URLs. Strips ads, navigation, and boilerplate — returns just the content.
Accepts up to 10 URLs in a single call. Multiple URLs are fetched in parallel server-side.

Flags

FlagDescription
--format <format>Output format: markdown (default), html, or json
--linksInclude extracted links from the page
--image-linksInclude extracted image URLs
--per-url-timeout-ms <milliseconds>Per-URL timeout budget, from 1 to 110000 milliseconds
--prettyHuman-readable output

Output

When --links or --image-links are set, each result also includes:

Examples


tinyfish agent run

Execute a browser automation. By default the output streams as newline-delimited JSON — one object per event.

Flags

FlagDescription
--url <url>Target URL to automate (required)
--syncWait for the run to complete and return the full result
--asyncSubmit only — return the run_id immediately without waiting
--output-schema <json>Inline JSON Schema object for structured output
--output-schema-file <path>Path to a JSON file containing the output schema
--browser-profile <profile>Browser profile for execution (lite or stealth)
--max-steps <n>Maximum tool-call steps before stopping (1-500)
--mode <mode>Agent behavior mode (default or strict). strict is in beta — join the beta to use it (contact support to enable it).
--session-id <uuid>Caller-provided UUID for idempotency / parallel-safe --sync
--use-vaultInject vault credentials into the run. Requires a connected vault provider.
--credential-item-id <id>Scope the run to a specific credential item by itemId (repeat for multiple; from tinyfish vault item list). Requires --use-vault. Omit to use all enabled items.
--prettyHuman-readable output instead of JSON

Output modes

One JSON object per line as events arrive. Use this when you want live progress or are piping to another tool.
The default stream also emits STREAMING_URL events (the live browser view) and periodic HEARTBEAT keep-alives.
Pressing Ctrl+C during a streaming run cancels the run server-side before exiting.

tinyfish agent run list

List recent runs.

Flags

FlagDescription
--status <status>Filter by status: PENDING, RUNNING, COMPLETED, FAILED, or CANCELLED
--limit <n>Number of runs to return (default 20, max 100)
--cursor <cursor>Pagination cursor from a previous response
--prettyHuman-readable output

Examples


tinyfish agent run get <run_id>

Get the full details of a specific run.
Returns the complete run object including status, result, and metadata.

Flags

FlagDescription
--prettyHuman-readable output

Example


tinyfish agent run steps <run_id>

Print the step-by-step trace for a run.

Flags

FlagDescription
--prettyHuman-readable output

tinyfish agent run watch <run_id>

Poll a run and print new steps as they arrive. Exits when the run reaches a terminal status.

Flags

FlagDescription
--timeout <ms>Stop polling after this many milliseconds (default 1800000)
--interval <ms>Poll interval (default 3000, min 500)
--prettyHuman-readable output

tinyfish agent run cancel <run_id>

Cancel a run that is PENDING or RUNNING.

Flags

FlagDescription
--prettyHuman-readable output

tinyfish agent batch run

Submit a batch of runs from a CSV file. The CSV must have url and goal columns.

CSV format

Flags

FlagDescription
--input <file>Path to CSV file (required)
--prettyHuman-readable output

Output

Batches are tracked locally in ~/.tinyfish/batches.json. Use batch list and batch get to check progress.

tinyfish agent batch list

List all locally tracked batches.

Flags

FlagDescription
--prettyHuman-readable output

Output


tinyfish agent batch get <batch_id>

Get run results for a batch, including per-run status and extracted data.

Flags

FlagDescription
--prettyHuman-readable output

Output


tinyfish agent batch cancel <batch_id>

Cancel all runs in a batch.

Flags

FlagDescription
--prettyHuman-readable output

Output


tinyfish browser session create

Spin up a remote browser instance. Returns a CDP (Chrome DevTools Protocol) WebSocket URL for programmatic control.

Flags

FlagDescription
--url <url>Navigate to a URL after session creation
--prettyHuman-readable output

Output

Use the cdp_url to connect with Playwright, Puppeteer, or any CDP-compatible client. The base_url can be used for HTTP-based interactions with the session.

Examples


tinyfish profile create

Create a new Browser Context Profile.

Flags

FlagDescription
--name <name>Profile name (required)
--prettyHuman-readable output

Output

Examples


tinyfish profile import-cookies

Extract cookies from your local Chrome install and upload them to a profile. This lets TinyFish reuse your existing logged-in session without you having to log in again during a run. Use --domain to import cookies for a single site, or --all-domains to import every domain in your Chrome profile in one shot. Exactly one of these flags must be provided — they are mutually exclusive.
Supported on macOS and Linux. Requires Google Chrome to be installed.

Flags

FlagDescription
--domain <domain>Domain to import cookies for, e.g. github.com. Mutually exclusive with --all-domains.
--all-domainsImport cookies for every domain in your Chrome profile. Mutually exclusive with --domain.
--profile-id <id>Profile to import into. If omitted, a new profile is created automatically.
--prettyHuman-readable output

Output

Examples

After importing, pass the profile ID to tinyfish agent run via --use-profile or include use_profile: true + profile_id in your API request to reuse the session.
See Browser Context Profiles for the full workflow.

tinyfish vault connection add

Connect a credential provider (1Password or Bitwarden). Credentials are validated immediately — if they’re invalid, the command fails. Provider secrets are read from environment variables, never passed as flags, so they stay out of shell history.

Flags

FlagDescription
--provider <provider>1password or bitwarden (required)
--client-id <id>Bitwarden client ID (required for Bitwarden; format user.xxxx-xxxx)
--server-url <url>Self-hosted Bitwarden server URL
--prettyHuman-readable output

Environment variables

VariableDescription
TINYFISH_VAULT_TOKEN1Password service account token (starts with ops_)
TINYFISH_VAULT_CLIENT_SECRETBitwarden client secret
TINYFISH_VAULT_MASTER_PASSWORDBitwarden master password

tinyfish vault connection list

List all connected vault providers and their status.

Flags

FlagDescription
--prettyHuman-readable output

tinyfish vault connection remove <connectionId>

Disconnect a provider. All cached credentials are removed immediately; runs already in progress are not affected.

Flags

FlagDescription
--prettyHuman-readable output

tinyfish vault item sync

Pull the latest credentials from all connected providers. Run this after adding new credentials or changing passwords in your vault.

Flags

FlagDescription
--prettyHuman-readable output

tinyfish vault item list

List all synced credential items. Each item has an itemId — that’s what you pass to --credential-item-id on agent run.

Flags

FlagDescription
--prettyHuman-readable output

Examples

See Vault & Credentials for provider setup instructions and troubleshooting.

Auth commands

CommandDescription
tinyfish auth loginOpen the API keys page and save a key interactively
tinyfish auth setRead an API key from stdin and save it
tinyfish auth statusCheck whether a key is configured and where it comes from
tinyfish auth logoutRemove the saved API key