Plain
Plain built an autonomous support agent, and has now built the controls to trust it.
A side-by-side editorial comparison of Frill and LiveAgent — release velocity, themes, recent moves, and the top alternatives to consider.
Frill is turning its feedback board from a destination into something machines read on demand.
The MCP server is the centrepiece: ten tools, one API key, no configuration, callable from Claude, ChatGPT, Cursor or Intercom Fin, and included on every plan rather than gated. The Chrome extension follows the same instinct from the capture side — grab an idea wherever you spot it, and collapse a long help desk thread into a clean idea with an AI summary in one click. Behind both sits the June groundwork: a public npm SDK, multiple API keys, webhook handling, and MCP in public beta.
LiveAgent is spending this window on scale defects, not on the AI surface it built in spring
Every release in the last six weeks is a build-stamped fix batch, and the failures being fixed are the kind that show up at tenant scale: a background ticket-search job crashing repeatedly on a deleted message record, a Facebook ticket poisoning its reindex batch forever with no dead-letter, a purge task full-scanning conversations fleet-wide, Microsoft shared-mailbox OAuth timing out on tenants with thousands of addresses. Work is being cut on two branches at once, 5.67.7 and 5.66.5, with the important fixes backported. Nothing in this window touches the MCP server or AI agent routing shipped in May and June.
The MCP server is the centrepiece: ten tools, one API key, no configuration, callable from Claude, ChatGPT, Cursor or Intercom Fin, and included on every plan rather than gated. The Chrome extension follows the same instinct from the capture side — grab an idea wherever you spot it, and collapse a long help desk thread into a clean idea with an AI summary in one click. Behind both sits the June groundwork: a public npm SDK, multiple API keys, webhook handling, and MCP in public beta.
Frill is systematically removing the requirement to be in Frill. Capture moves to wherever the conversation already happens; retrieval moves to whatever assistant the team already uses. That reframes the product as the feedback store rather than the feedback interface, and the pricing decision reinforces it — putting MCP on every plan optimises for the corpus being queried, not for upsell. The earlier enterprise work (EU hosting, EntraID SSO, CSP) is what makes that store trustworthy enough to hold the data.
Expect write paths to follow the read paths — an assistant that can already ask what users want will next be expected to file, merge and status ideas — and expect more capture surfaces beyond the browser extension.
Every release in the last six weeks is a build-stamped fix batch, and the failures being fixed are the kind that show up at tenant scale: a background ticket-search job crashing repeatedly on a deleted message record, a Facebook ticket poisoning its reindex batch forever with no dead-letter, a purge task full-scanning conversations fleet-wide, Microsoft shared-mailbox OAuth timing out on tenants with thousands of addresses. Work is being cut on two branches at once, 5.67.7 and 5.66.5, with the important fixes backported. Nothing in this window touches the MCP server or AI agent routing shipped in May and June.
This is the consolidation phase after a fast AI push: the MCP server, AI Agent Work Distributor and per-agent auth tokens landed in spring, and the engineering attention has since moved to durability and to the integrations that feed the helpdesk. Maintaining a parallel 5.66 branch with backports indicates customers who will not move versions quickly, which slows how fast any of the newer AI surface actually reaches the base.
Expect the fix cadence to continue on both branches, with the next feature-carrying release most likely extending the existing MCP tool set rather than opening a new AI surface.
Other Support products tracked by Sparkpulse, ranked by recent ship velocity. Each card links to a full editorial trajectory and lets you pivot into a head-to-head comparison with either Frill or LiveAgent.
Plain built an autonomous support agent, and has now built the controls to trust it.
Sleekplan rebuilt itself around AI triage, then put the feedback board inside ChatGPT.
Comm100 publishes steadily about AI support — none of it is Comm100 shipping anything.
respond.io is layering an agent-authoring copilot and consumption billing on top of its omnichannel inbox.
Hatz sells through MSPs, and its roadmap has started to serve the partner more than the end user
A teaser-only content feed making one argument: AI in the contact center fails on knowledge, not models.
See all Frill alternatives → · See all LiveAgent alternatives →
Latest ship moves from both products, interleaved chronologically. ⚡ = editorial spark.
They serve adjacent needs but don't currently overlap on shipped themes. Frill is currently shipping more aggressively (velocity 6.3 vs 5.0), with 1 editorial sparks in the last 30 days against 0. See the at-a-glance table above for a side-by-side breakdown of velocity, recent sparks, and editorial themes.
Sparkpulse doesn't pick a winner — we score release velocity, not feature parity. Frill is currently shipping more aggressively (velocity 6.3 vs 5.0), with 1 editorial sparks in the last 30 days against 0. For your specific use case, the alternatives sections above list other Support products to evaluate alongside.
Top Frill alternatives in Support are ranked by recent ship velocity. Browse the "Frill alternatives" section above for the current picks, or visit /alternatives/frill for the full list with editorial commentary on each.
Top LiveAgent alternatives in Support are ranked by recent ship velocity. Browse the "LiveAgent alternatives" section above for the current picks, or visit /alternatives/liveagent for the full list with editorial commentary on each.