Dust vs Ably
Side-by-side trajectory, velocity, and editorial themes.
Dust is widening the agent-platform surface: multimodal tools, enterprise audit, model breadth.
Dust is shipping at a fast clip on three fronts that together define a serious agent platform: model breadth (Gemini 3.5 Flash, Grok 4.3, refreshed Anthropic lineup), agent capability (MCP tools can now return images the agent can actually see, plus context compaction for long runs), and enterprise readiness (workspace audit logs streamable to Datadog, Splunk, or any HTTPS sink). Integrations are getting versioned upgrades on the side (Asana MCP v2, Gmail labels and archive). The product is moving from 'chat with an agent' toward 'run agents in production with observability and multimodal I/O.'
Two clear directions: deeper enterprise GTM via SIEM-grade audit, and a more capable agent runtime that can see, remember, and act inside third-party SaaS. The MCP-image release in particular treats Model Context Protocol as a real I/O surface rather than a text-only RPC, which is where the broader MCP ecosystem is heading. Frequent model rotations suggest Dust is positioning as model-agnostic infrastructure rather than locking into one provider.
Next moves likely lean into the same arc: more MCP integrations with action verbs (write/delete/transition states), expanded multimodal returns (audio, structured documents), and finer-grained admin controls layered on top of the audit foundation - tool-usage policies, per-agent egress rules, or approval workflows.
Ably pivots its developer surface toward AI agents as first-class consumers
Ably's developer surface has been visibly pivoting toward AI agents over the last six weeks. The CLI hit v1.0 with structured JSON output envelopes, hint fields for self-healing agents, and unified --force semantics. Two weeks later v1.1 added `ably init`, which installs the CLI plus Agent Skills bundles directly into Claude, Cursor, Windsurf, and VS Code in one step. A new AI Transport SDK lands separately, with end-to-end tool-calling and React providers for the Vercel AI SDK.
Ably is building two integration stories in parallel — agents as CLI users (via Agent Skills) and agents as transport consumers (via the AI Transport SDK). Together they bracket the workflow: agents read live data through the SDK and operate Ably itself through the CLI. The traditional client-library stream (JS, Python, Laravel) continues at steady pace in the background, with LiveObjects formally graduating to GA on the protocol-6 update.
Expect the AI Transport SDK to add tool-call adapters for at least one more AI runtime beyond Vercel — most likely the OpenAI Agents SDK or Anthropic's Claude Agent SDK — and Agent Skills coverage to expand to additional AI clients (Cline, Aider, Continue). Watch for the LiveObjects API to surface inside the chat/AI surfaces.
See more alternatives to Dust →
See more alternatives to Ably →