Back to home

Product Updates

New capabilities, experience improvements, and important fixes.

FastClaw foundation upgrade and a rebuilt multi-Agent workbench

The workbench now uses the updated native FastClaw interface and runtime, adapted to our accounts, credits, and seller workflows. Capabilities previously selected through buttons in one assistant are now separate General Assistant, Market Research, and Analysis Report Agents in the sidebar. Users can also create private Agents. The layout, conversation flow, and file delivery have been rebuilt, with follow-up experience fixes now live.

Affected areas

  • FastClaw interface and runtime
  • Workbench layout and site branding
  • General Assistant, Market Research, and Analysis Report
  • External data queries through MyLevis MCP Hub
  • Private Agents and access permissions
  • Conversation history, tool activity, and disconnect recovery
  • Credit disclosure, scope approval, and report files

What changed

  • General Assistant now connects to MyLevis MCP Hub and can query connected SIF, SellerSprite, and Sorftime tools for review summaries, reviews, keywords, and traffic as needed. Up to three queries are allowed per turn and included in the current ordinary-chat credit price. Tool activity shows query status without exposing service credentials. Market Research and Analysis Report retain their dedicated workflows.
  • The workbench follows FastClaw’s native layout and interactions, with colors, fonts, icons, and branding aligned with our account and settings pages. Site capabilities stay in an adapter layer to make future upstream updates easier to follow.
  • Choose General Assistant, Market Research, or Analysis Report in the sidebar. Each Agent retains its own conversation ownership and supports follow-up questions in the relevant conversation.
  • Native conversation and file panels on the right replace the legacy report-history entry and permanent plan panel. Plans requiring approval and execution status appear in the conversation. Legacy conversations are not migrated into the new workbench.
  • FastClaw’s native Agent creation is now available: provide a name and purpose to create a private Agent. Creation uses no website credits; conversations follow the existing credit rules.
  • Edit the name and purpose of your own Agents, or delete them after confirmation. Running sessions block edits and deletion; built-in Agents remain read-only. Deletion removes the Agent’s runtime conversations and files while retaining credit records.
  • Private Agents, conversations, and files are isolated by user. System administration, model configuration, team permissions, and other control-plane settings are not exposed to users.
  • Native tool activity is restored, with collapsible details and saved activity available after refresh. Commands, internal configuration, and other sensitive details are filtered before display.
  • Disconnect and refresh recovery now uses the runtime record for the exact task to recover results, tool activity, and reports, reducing false failures while long tasks are still running.
  • The composer shows the balance and billing policy before sending, distinguishing ordinary chat, analysis execution, and Market Research reservations. Scoping and execution within an already reserved investigation are not charged again.
  • Market Research shows its approval action only when the service has actually prepared a scope for confirmation. Missing information is clarified in the conversation before formal research starts.
  • Completed analysis reports appear in the native Files panel with readable dates and sequence numbers, direct preview, and download. Viewing or downloading a saved report uses no credits.
  • New reports and prompts use readable source references and explanations, reducing internal identifiers in the reading experience. Workbench buttons, composer controls, and reading details have also been refined.

Credit pricing has been upgraded

Credits are now priced around successfully delivered seller outcomes, with clearer value tiers for quick chat, tool-assisted work, and complete Market Research.

Affected areas

  • Credit packs
  • Task billing
  • New-user trial
  • Referral rewards

What changed

  • Quick chat uses 2 credits, tool-assisted work 12, and complete Market Research 100.
  • New users receive 20 credits valid for 7 days.
  • New packs are $15 / 100, $39 / 360, and $99 / 1,000 credits.
  • Both users receive 10 credits after verification, and the purchase reward is now 10%.

Referral rewards are now available

Signed-in users can now share a personal referral link. Both users receive credits after the friend signs up by email and verifies their address, with an additional reward when the friend purchases credits.

Affected areas

  • Referral links
  • Email signup and verification
  • Credit rewards

What changed

  • The inviter and invited user each receive 15 credits after email verification.
  • There is no limit on invitations; each invited account can bind to one inviter.
  • The inviter receives 30% of the credits in each package the friend purchases.

Beta experience fixes

A few early beta issues were fixed and refined. The workspace now includes an attachment entry and a right-side context panel, and support ticket screenshot uploads use a more stable storage path so users can share visual context during testing.

Affected areas

  • Agent workspace
  • Attachment uploads
  • Ticket screenshot feedback

What changed

  • Attachments can be selected before sending and are shown in the conversation.
  • The right-side workspace panel now summarizes the current conversation, attachments, and latest reply.
  • Fixed ticket screenshot uploads that could fail in some environments.

Improved long-running task stability

Long-running Agent tasks are now more stable. Some market research, Listing generation, and similar tasks could still be processing while the page showed an early timeout or failure. Long tasks now have more reliable connection handling and timeout behavior while work is running.

Affected areas

  • Market research tasks
  • Listing generation tasks
  • Other Agent tasks that need more time to complete

What changed

  • Long tasks wait more reliably.
  • The page is less likely to report failure while work is still in progress.
  • Longer research, organization, and generation tasks are safer to run.

MVP beta release and connectivity testing

The Cross-Border AI Vision MVP beta was released and completed the first round of user-side connectivity testing. Core flows including signup, email verification, trial credits, Agent conversations, follow-up questions, and history were verified for early-user testing.

Affected areas

  • Account signup and email verification
  • Trial credits
  • Agent workspace
  • Conversation history and follow-up questions

What changed

  • Users can sign up by email and use the workspace after verification.
  • New users receive trial credits for testing Agent conversations.
  • Tasks can be started in the workspace and continued from history.
  • This stage focuses on early testing and manual support.