{"version":1,"type":"story","url":"https://digestai.news/story/perplexity-details-brain-memory-system-and-agent-api-testing-methods","json":"https://digestai.news/story/perplexity-details-brain-memory-system-and-agent-api-testing-methods.json","markdown":"https://digestai.news/story/perplexity-details-brain-memory-system-and-agent-api-testing-methods.md","slug":"perplexity-details-brain-memory-system-and-agent-api-testing-methods","headline":"Perplexity details Brain memory system and Agent API testing methods","summary":"Perplexity published a technical guide on November 2, 2026, explaining its **Brain** memory architecture for **Perplexity Computer**, a system designed to store and organize agent interactions across three layers: raw transcripts, distilled notes, and a linked knowledge wiki. Brain uses token-level memory outside the model, allowing agents to retrieve and update context without altering learned weights. Background agents called **Dream** process daily sessions, summarize records, and revise the wiki via Git-like diffs, reinforcing valid information and marking outdated entries. Perplexity reported **9.3% higher correctness** and **8.9% better recall** in 30-day tests, alongside a **15% token reduction** compared to baseline agents. In a separate benchmark, an agent using Brain-based tools scored **0.627** on financial-document questions, outperforming semantic search (**0.37** on the same corpus).","keyPoints":[],"whyItMatters":"Brain’s architecture could redefine how AI agents retain and update context, reducing token costs and improving task accuracy. For developers, the distinction between Brain (self-maintaining) and the Agent API (manual control) clarifies trade-offs. The guide’s benchmarks and testing frameworks may become industry standards for evaluating memory-dependent agent performance.","category":{"slug":"agents","name":"Agents & Tools","url":"https://digestai.news/category/agents"},"entities":{"companies":["Perplexity"],"models":["Perplexity Computer","Brain"],"people":[]},"firstPublishedAt":"2026-11-02T16:21:03Z","updatedAt":"2026-11-02T16:21:03Z","sourceCount":1,"hasPrimarySource":false,"sources":[{"outlet":"Unite.AI","title":"Perplexity Details Brain Memory Architecture and Agent API Testing","url":"https://unite.ai/perplexity-details-brain-memory-architecture-and-agent-api-testing","publishedAt":"2026-11-02T16:21:03Z","type":"press","primary":false,"lead":true}],"sourceNotes":null,"discussions":[],"thread":null,"cite":{"text":"Digest AI, \"Perplexity details Brain memory system and Agent API testing methods\", 2 November 2026, https://digestai.news/story/perplexity-details-brain-memory-system-and-agent-api-testing-methods","publisher":"Digest AI","title":"Perplexity details Brain memory system and Agent API testing methods","datePublished":"2026-11-02T16:21:03Z","url":"https://digestai.news/story/perplexity-details-brain-memory-system-and-agent-api-testing-methods"},"generatedBy":"Written by Digest AI's editorial model from the linked sources; the sources are the record.","license":"Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse"}