Metaphor
When customers ask questions your training data does not cover, your AI agent reaches out to the web through Metaphor's neural search engine. It fetches relevant content, extracts the information needed, and delivers accurate, up-to-date answers grounded in real sources.
Metaphor's neural search understands meaning, not just keywords. Your AI agent retrieves and processes web content to answer questions with sourced, accurate information.
Metaphor
See how AI agents use Metaphor to go beyond static training data, pulling fresh information from the web to answer questions that would otherwise hit a dead end.
A prospect asks 'How does your product compare to Competitor X?' Your AI Agent fetches relevant review pages and comparison articles through Metaphor, extracts the key differentiators, and presents a balanced, sourced comparison. The prospect gets a helpful answer instead of 'I can only speak about our product.' Credibility builds trust that converts.
A customer asks about a recent regulatory change affecting your industry. Your AI Agent retrieves the latest articles and official documents through Metaphor's content API, extracts the relevant details, and explains how the change affects them. Information that is days old, not months, delivered in real-time conversation.
A B2B customer wants to understand best practices for implementing your product category. Your AI Agent fetches authoritative blog posts, whitepapers, and documentation through Metaphor, synthesizes the key takeaways, and shares them with source links. The customer gets expert-level guidance backed by real references, elevating your support beyond generic FAQ responses.
Metaphor
FAQs
Metaphor uses an embeddings-based neural approach that understands meaning rather than matching keywords. It was specifically designed for AI systems to use, returning structured, parsed content rather than links. The results are semantically relevant, meaning the agent finds contextually appropriate content even when the query uses different terminology than the source.
Metaphor has indexed a large portion of the web and can retrieve content from most publicly accessible pages. The Get Contents endpoint accepts both Metaphor document IDs and direct URLs. Some dynamically rendered or paywalled content may not be fully accessible. The agent receives parsed text and metadata for available pages.
The agent handles this gracefully. If no relevant content is found, the agent falls back to its training data or lets the customer know it could not find specific external information on that topic. It can also suggest the customer rephrase their question or offer to connect them with a human agent.
No. Content is fetched in real-time from Metaphor during the conversation and used only to formulate the current response. Web content is not cached or stored in Tars systems. Each request retrieves fresh data from Metaphor's index.
You can control this precisely. The Get Contents endpoint accepts startChar and endChar parameters to retrieve specific text ranges. For full documents, the entire parsed text is returned. For multiple documents, up to 1,000 IDs can be processed in a single API call, though typically the agent retrieves only what is needed for the conversation.
Yes. Metaphor rebranded to Exa. The underlying neural search technology is the same. The Tars integration uses Metaphor's content retrieval API, which remains fully functional. Both names refer to the same search engine and company.
The Get Contents endpoint retrieves content for specific document IDs or URLs you provide. By controlling which IDs the agent uses, you effectively control the sources. For use cases where source restriction is critical, configure your agent's gambits to only pass vetted document IDs or approved URL domains.
Metaphor's index is regularly updated, but there may be a lag between a page changing and the index reflecting the update. For content that changes daily, the agent retrieves the most recent indexed version. For time-sensitive information, pairing Metaphor with other real-time data sources provides the most current answers.
Don't limit your AI Agent to basic conversations. Watch how to configure and add powerful tools making your agent smarter and more functional.
Privacy & Security
At Tars, we take privacy and security very seriously. We are compliant with GDPR, ISO, SOC 2, and HIPAA.
Email - sales@hellotars.com
.png)
%201.png)
%201.png)
