Enterprise Brain: Context That Makes Agents Better

AmpUp launched Enterprise Brain, a versioned memory layer that connects existing enterprise context graphs and catalogs to AI coding agents including Claude Code, Codex, Gemini CLI, Cursor and Copilot, giving them shared memory that learns from graded outcomes while keeping source permissions and audit trails in customer-controlled cloud tables. In AmpUp's own evaluations, root-cause accuracy on the 253-task ORCA-bench rose from 6.3% to 35.2% at 25% lower cost ($0.032 per task), CRM question accuracy on CRMArena-Pro climbed from 57.6% to 76.7%, and LongMemEval-S improved 21.3 points to 78.7% at $0.0012 per question. Smaller gains came on code review (17.1% to 20.3% on 688 real cases, but 43% more expensive at $0.024 per review) and DolphinBench (62.9% to 72.3%). The company claims best-in-class memory based on these named baselines, though it notes broader leaderboard standing has not been independently verified. The pitch matters because agent vendors are competing on persistent, governed enterprise memory rather than model quality alone; watch for independent benchmark replication and whether the versioned-trajectory records translate into measurable training-data advantages.