← Back to all sparks
Front logo

Front

SUPPORTCOLLABCOMMS
Velocity5.0

Collaborative inbox and customer communication platform

Front stops treating AI as a feature and starts managing agents as staff

ai-agentsagent-evaluationmcpshared-inboxaccountssupport-qa
◆Current state
Front has moved from auditing AI replies to running AI agents inside the shared inbox. Since 6 October, teams can bring in Autopilot agents or external agents built with Claude, Codex, Sierra or another provider, with the ownership and context a human teammate gets. Compass scores agent replies, the October digest adds agent evaluation and testing tools, and the MCP server was scheduled for all plans on 6 October. Accounts and Copilot routines add a second thread aimed at account management rather than ticket handling.
◆Where it's heading
The audit tooling Front built through the summer now has a larger job: if external agents run in the inbox, Front's value sits in coordination, ownership and measurement rather than in its own model. Guided handoffs, evaluation and testing all fit that reading. The MCP server points the same way from the other end, letting outside assistants read and act on Front conversations. Accounts and churn-risk routines push the product from support toward the wider customer team.
◆Prediction
Expect more evaluation and handoff controls for external agents, since testing and evaluation are the newest parts of the stack. The digests give too little detail to say how far Accounts will go as a customer-management surface.

◆Recent moves

  1. 15d ago

    What’s New in Front, October 2026: AI agents, proactive account tools, richer reporting

    The October digest fills in the agent story: guided handoffs between people and agents, evaluation and testing tools, and Accounts with AI summaries. It also adds token-level API analytics and a Modem integration, which keeps the release as much about measurement as about AI.

  2. 1mo ago

    Front brings Autopilot and external agents into the shared inbox

    Front opens the inbox to custom Autopilot agents and to external agents built with Claude, Codex, Sierra and others, with Compass scoring their replies. This is the turn from AI as a feature to agents as managed teammates, and it reframes the audit work of earlier months.

  3. 1mo ago

    What’s New in Front, September 2026: MCP, smarter AI workflows, and service signals

    September sets up the agent launch: the MCP server is announced for all plans on 6 October, Admin Copilot recommends assignment rules, and Autopilot now moves ticket statuses itself. ClickUp joins the knowledge sources, and Smart CSAT scores become usable in rules.

  4. 2mo ago

    What’s New in Front, August 2026: trusted AI, clearer QA, customer journey insights, and new integrations

    August extends the audit-the-AI arc with an AI replies hub that exposes which knowledge sources fed a reply and allows inline correction, plus GitBook as a source. The larger share of the month is QA maturity: Smart QA and CSAT now evaluate every teammate on a conversation, with analytics filters and conversation-level exports.

  5. 3mo ago

    What’s New in Front, July 2026: redesigned mobile app, smarter AI, new integrations

    July's digest pairs the redesigned mobile app with governance plumbing — Admin Copilot audit log actions, topic management, default channel signatures. It sets up the October mobile cutover that the August digest then dates.

  6. 4mo ago

    What's New in Front — June 2026 (AI file uploads, 4 new integrations, early access features)

    A duplicate capture of the June 2026 digest — byte-identical body to the entry above it, differing only in how the headline is punctuated. No separate release behind it.