Fibery vs Zoho Sprints

Fibery is shipping weekly product updates at high velocity, with AI agent capabilities as the organizing thread. The September 17 release rebuilt the AI agent from scratch — replacing a multi-tier router/sub-agent architecture with a single streaming instance supporting dynamic tool loading and a model picker that spans Haiku through Opus 5.5, GPT 6, Gemini, and Astra 6. The same release added OAuth Applications, which unblocks enterprise MCP client connections without filing a support ticket. Custom Apps received a major platform revision the week before, with breaking API changes signaling a stable, long-term surface. Whiteboard AI access, batch undo designed explicitly for AI error recovery, and per-source change tracking (distinguishing UI, automation, MCP, and AI changes) all shipped in this window.

Fibery is building end-to-end agentic work management: custom apps authored via MCP, AI that writes automations, batch undo that anticipates AI errors at scale, and OAuth enabling any external AI client to connect without support friction. Each release tightens the oversight loop — users can see what an agent touched, undo it in bulk, and trace it to a specific tool. The 1.5 people on performance work (with 'many parallel agents', per their own release note) suggests the team is using the product to build the product. The weekly cadence points to rapid PMF iteration on the AI-native positioning.

Whiteboard spatial AI improvements are explicitly flagged as near-term — they noted that spatial operations require top models. Field usage tracking will expand to cover views, validation rules, and access templates (listed as upcoming). The AI agent's model picker and the OAuth surface together suggest a push toward bringing third-party LLM providers in as first-class options rather than keeping Fibery's own AI credits as the primary path.