PostHog vs Neo4j
Side-by-side trajectory, velocity, and editorial themes.
PostHog is wiring itself into the MCP ecosystem while shoring up mobile-SDK feature parity.
PostHog continues its weekly grind, but the May releases cluster around two themes: an MCP toolchain (alerts to Slack and webhooks, SDK Doctor, mode selection via header) and LLM analytics BYOK providers (Together AI, Azure OpenAI). At the same time the mobile teams are filling in iOS and Android session-replay controls, rage-click detection, and survey delays that previously only the web SDK had.
The shape of PostHog's surface keeps widening rather than deepening: more LLM-vendor coverage in the analytics product, more MCP-tooling so AI agents can read and act on PostHog data, more parity across SDKs. Less obvious is which surface becomes the headliner; right now Conversations, Logs, Experiments, and Client Libraries are all shipping into a single weekly digest with comparable weight.
Expect MCP integration to keep expanding from peripheral utilities into the core insights and alerting paths, with PostHog positioning itself as the analytics endpoint AI agents read from when reasoning about product usage. Mobile SDK parity work should compress in the next month or two as the gap with the web SDK closes.
neo4j-cli ships explicitly for AI agents — Neo4j makes its 'AX' bet concrete.
Neo4j is shipping in three lanes simultaneously: developer/agent surface (the new neo4j-cli covering Aura management, Cypher, and ops, designed for human, developer and agent consumption), Aura cloud capacity and ops (2TB high-memory GCP instances, inactive-member pruning, tighter password policy), and graph analytics maturation (project-level ML model persistence in AGA, Lakehouse export from Microsoft Fabric, Cypher 25 GQL features). Dashboards and Explore are gaining interactivity in parallel.
The arc is toward treating AI agents as a first-class user of the platform, not an integration consumer. Calling out 'AX' alongside DX/UX in the CLI announcement is unusual — most database vendors are still adding MCP servers or chat assistants. Coupled with the GenAI token functions in the April Aura release and AGA's model persistence, Neo4j is consolidating the 'graph as memory substrate for AI agents' position it's been telegraphing for two years.
Likely next: an MCP server fronting the same surface as neo4j-cli, deeper GenAI-native primitives in Cypher 25 (vector ops, embeddings as first-class types), and continued Aura capacity climbs to support larger graph-RAG workloads. Microsoft Fabric integration will probably extend further given the bidirectional Lakehouse work.
See more alternatives to PostHog →
See more alternatives to Neo4j →