PhoenixAI Launches Contextbase for AI Agents
The context infrastructure stores semantics, knowledge, memory, traces, and evaluations in a single system, unveiled at PhoenixAI's NEXT QUERY 2026 summit.
Updated on Nov. 5, 2026 in Artificial Intelligence
PhoenixAI launched Contextbase, a context infrastructure for AI agents, at its NEXT QUERY 2026 Data + AI Infrastructure Summit. The system stores semantics, knowledge, memory, traces, and evaluations in one place and will enter private preview on November 4.
Why it matters
Agents only know what is in their context, and in most current stacks that context is split across separate tools that each keep their own copy and permissions. Consolidating it into one system aims to reduce that fragmentation for developers building agent applications.
Contextbase is built into the PhoenixAI database, with context living on the same cluster as business data so one query can reach all of it under one set of permissions. Traces record every model call and tool call in a run, including prompts, outputs, tokens, and cost, though no throughput or latency figures were disclosed.
The players
PhoenixAI
PhoenixAI is a Menlo Park, California-based company that develops data and AI infrastructure products, including the PhoenixAI database.
The details
Semantic views of entities, dimensions, and metrics are defined by the business in Semantic SQL, while knowledge records are typed, versioned, and linked to each other. Teams can request a free trial at phoenixdata.ai, and PhoenixAI is also building ObserveStack, an application on Contextbase for tracing and evaluating agents.
Timeline
November 4, 2026: Contextbase becomes available in private preview.
November 5, 2026: PhoenixAI announces the Contextbase launch at NEXT QUERY 2026.
The Tech Race
Contextbase extends the retrieval-augmented generation (RAG) pattern, which pairs a vector store with a separate memory service, by folding semantics, memory, traces, and evaluations into one system. It positions PhoenixAI as a competitor to the fragmented multi-tool stacks that dominate agent infrastructure today.
Developers building AI agents may be able to consolidate separate semantic layer, vector store, memory, and observability tools into one system governed by a single set of permissions. Teams can request a free trial at phoenixdata.ai ahead of the November 4 private preview.
The takeaway
Organizing agent context in one governed place rather than across several tools may simplify how teams build and audit AI agents. Teams evaluating the approach can follow the November 4 private preview to see how it performs in practice.
What happens next
Contextbase becomes available in private preview on November 4, 2026, and teams can request a free trial at phoenixdata.ai.
Further reading
For broader background, explore Artificial Intelligence coverage from Menlo Park.
More information
Request a free Contextbase trial from PhoenixAI.
Source note: This article includes information reported by HPCwire.
