Developer Platforms
Zoho Makes Catalyst Agent-Ready but Keeps Production Promotion Human
Zoho announced Catalyst Agent Skills, MCP support, a noninteractive CLI, and coding-assistant integrations on September 2. Its product page says agents use scoped collaborator permissions, destructive noninteractive commands are disabled, logs preserve tool activity, and production promotion remains manual. Zoho's completion-rate table is first-party and lacks a full public evaluation protocol.
Citation-ready: Zoho announced on September 2, 2026, that Catalyst now supports Agent Skills, MCP, and a noninteractive CLI while retaining scoped agent permissions and manual promotion to production.

What happened and why it matters
No. Zoho's stated control model separates development from production, scopes the agent's collaborator identity, disables destructive commands in noninteractive mode, logs activity, and keeps production promotion manual.
Zoho launch release, product page, and updated documentation
Primary reference: Zoho: Catalyst Agent Skills and agent-ready platform launch. Kaleido Field checked the event date and the article's attributed facts against this source.
| Source date | September 2, 2026 |
|---|---|
| Checked by Kaleido Field | September 3, 2026, 09:20 CST |
| Source function | current developer-platform analysis separating Agent Skills, MCP, noninteractive CLI behavior, scoped permissions, disabled destructive commands, audit logs, manual promotion, first-party task results, and production evidence |
Noninteractive should mean explicit, not silent
Zoho says the CLI accepts required choices up front and stops when a needed value remains missing. That is a useful contract for agents because a terminal prompt cannot be mistaken for successful completion.
Each run should retain the requested task, skill version, model, CLI version, arguments, defaults, permissions, commands, tool calls, files changed, tests, exit status, and unresolved prompts.
Production remains a separate decision
Manual promotion preserves a review point between code generation and customer impact. Scoped collaborator permissions and disabled destructive commands reduce authority before that point, while logs help reconstruct what happened.
Teams still need branch protection, secret boundaries, dependency review, tests, infrastructure diff, approver identity, deployment receipt, health checks, rollback, and post-release monitoring around the platform controls.
Evidence boundary
Official product and documentation facts: named features and integrations, CLI motivation, up-front options, fail-fast missing input, separate environments, manual production promotion, scoped collaborator controls, disabled destructive commands, and platform and MCP logs. Zoho-reported evaluation: three model comparisons reaching 90% to 95% task completion with skills and MCP. Not established: public task set and scoring protocol, independent reproduction, security-test results, model parity, every command boundary, deployment quality, defect rate, or production outcome.
FAQ
What did Zoho add?
Agent Skills, MCP support, noninteractive CLI operation, and integrations including Claude Code and Codex.
Can the agent promote directly to production?
Zoho's product page says production promotion remains manual.
Are Zoho's completion results independent?
No. The comparison table is first-party and does not publish a complete external evaluation protocol.