Invite 10 friends.
Unlock 10M extra
tokens a day.

We’ve upgraded LangSearch around the things an AI agent needs most: useful context, fast responses and room to keep working. To celebrate, we’re giving builders a way to grow their limits for free.

One friend. A bigger daily allowance.

Share your personal invite link with a friend who hasn’t registered for LangSearch. When they sign up through your link and complete their first successful Web Search API call, you earn +1 RPS and +1,000,000 tokens per day.

With 10 qualifying friends, that adds up to 10 million extra tokens every day and 10 extra requests per second. These rewards increase your daily token ceiling; they aren’t a one-time bundle that disappears after you spend it. Your allowance resets at 00:00 UTC each day.

1. Copy your link. Open Invite friends in your Dashboard.
2. Share it with a builder. They must create a new account through that link.
3. Let them try search. Their first successful Web Search call qualifies the invitation. Rewards are checked automatically, normally within about a minute.

Each new friend can qualify once, self-referrals don’t qualify, and rewards stop after 10 qualifying friends. Track pending invitations, earned rewards and your daily token bonus directly in the Dashboard.

Better answers start with better search.

We’re building toward state-of-the-art search for AI: retrieving the right context quickly enough for an agent to act on it. Our latest reported results put LangSearch at 95.37% on SimpleQA and 100ms response latency.

Accuracy matters because irrelevant context can send an otherwise capable model in the wrong direction. Latency matters because agents often search several times before they finish a task. We’re improving both, so search can be part of the workflow instead of a pause in it.

These are LangSearch team-reported figures. The latency figure does not specify a percentile or test conditions, so it is not a promise that every request finishes in 100ms or a controlled comparison across providers. See our benchmark charts and measurement notes for the published comparison.

More control over the context you retrieve.

We’re focusing LangSearch on one product: Web Search for AI agents, with the controls to make the response useful in your application.

• Up to 50 results per request. Choose the breadth your task needs.
• Snippets or full webpage text. Set contents.text: true for text, or use contents.text.max_characters to control its length. Text is returned instead of snippets.
• Precise date filters. Use a freshness preset, a single date or a date range.
• Domain controls. Include trusted domains or exclude sources that don’t fit your task.
• API, MCP and Skill. Call the API directly, connect your tools to our official MCP server or give your agent the Skill installation prompt.

All of these supported features are included in the Free Plan. We’re no longer offering a separate Rerank API.

A new usage model. Still $0.

The old daily limit counted requests. The new model measures the text your search actually uses: the query as input tokens, and returned snippets or webpage text as output tokens. A short lookup and a request for 50 long pages no longer consume the same unit of allowance.

When webpage text is returned, we count that text without also charging for a snippet. Each response includes usage.input_tokens and usage.output_tokens, so your application can track usage as it runs.

Input and output token prices are currently zero. Token metering determines quota usage; it does not turn this update into a balance deduction. There’s no extra feature fee for full webpage text or domain and date filters.

The redesigned Plan & usage page shows tokens used today, the percentage of your daily allowance used, your next reset and your current RPS and TPM limits. All API keys on your account share those limits.

From 1 RPS to 5 RPS. Then grow it.

Previously, the standard starting limits were 1,000 requests a day and 1 request per second. New accounts now start at 5 RPS—five times the starting request rate—and daily usage is governed by tokens instead of a fixed 1,000-call ceiling.

There’s no fixed conversion from 1,000 requests to tokens: the number of searches your allowance supports depends on query length and returned content. RPS controls request frequency, TPM controls tokens per minute, and the daily token allowance controls total daily usage.

Existing accounts keep their saved RPS entitlement, including earlier increases from adding credits. Invitation rewards stack on top. For example, a new account starting at 5 RPS can reach 15 RPS after 10 qualifying invitations. Adding credits can also increase RPS; invitations are the way to earn the extra daily token allowance described here.

Bring a friend. Build something useful.

Have a teammate building a research assistant? A friend connecting their coding agent to the web? Send them your link, help them make their first successful search, and give your own agents more room to work.

Our direction hasn’t changed: The World Engine for AGI. This release makes that connection faster, more flexible and easier to build on.