• Updates

  • This update puts Session Key mode closer to the center of the AABC experience. In high-risk crypto workflows, the problem is not only how to move faster. It is how to move without falling into the usual loop of wallet switching, repeated confirmations, and fragmented trust. AABC now does a better job of turning risky actions into visible, structured, and reviewable execution paths, with Session Key mode reducing friction while keeping the experience easier to follow and trust.

    Added

    • Stronger Session Key mode support for conversation-first crypto execution

    • More audit-style execution guidance for sensitive crypto actions

    • Better visibility into what the agent is checking before and during execution

    Improved

    • User confidence when handing complex protocol actions to an agent through Session Key mode

    • The feeling of continuity between intent, execution, and result verification

    • Clarity around risky paths, approvals, and execution boundaries

    Fixed

    • Gaps between user expectation and visible execution progress

    • Weak trust signals in workflows that required more structure and reassurance

    • Moments where the system felt too opaque during sensitive execution

  • This update marks a bigger product shift: AABC is no longer best understood as a single agent experience. It is becoming a configurable agent system. Skills, knowledge, integrations, workflows, triggers, and runtime choices can now come together as part of how each agent works. That gives users a more powerful way to shape outcomes instead of just sending one-off prompts.

    Added

    • A more configurable agent workspace built around reusable operating layers

    • Live skill selection closer to the point of execution

    • Better alignment between agent setup, runtime choice, and current workflow intent

    Improved

    • The relationship between configuration and actual runtime behavior

    • The ability to treat agents as adaptable systems instead of fixed personas

    • Continuity between home entry, chat entry, and agent-specific surfaces

    Fixed

    • Gaps between saved configuration and per-run execution behavior

    • Inconsistent behavior across different entry points into the same workflow

    • Cases where live choices were lost before the run actually started

  • This update strengthens one of the most important trust layers in AABC: access belongs to the user who configured it. Skill readiness, external service access, and setup state now align much more closely with the account that actually owns the credentials. That makes the platform safer, less confusing, and much more credible in real use.

    Added

    • Stronger user-scoped access handling for services and skill setup

    • Clearer separation between public skill availability and private credential readiness

    • A more explicit setup flow for skills that require external keys or access

    Improved

    • Consistency between saved access, visible readiness, and actual execution availability

    • Trust in multi-account and multi-agent usage scenarios

    • User understanding of why a skill is ready, blocked, or still requires setup

    Fixed

    • Readiness leaking across different user accounts

    • Cases where backend-provided access looked globally available in the UI

    • Out-of-sync states where clearing saved access did not reset readiness correctly

  • This update moves AABC beyond “more tools” and into a more meaningful execution surface. Agents can now work with a broader layer of skills, protocol-facing knowledge, and service access patterns, making the system better suited for research, analytics, documentation, and crypto-native execution. The goal is not just tool count. It is turning more of the stack into something agents can actually use.

    Added

    • A broader skills layer for domain-specific operating knowledge

    • Stronger protocol-facing access patterns for service-based crypto workflows

    • Better routing between user intent, skills, and external action surfaces

    Improved

    • Agent selection of the most relevant skill or access path for a task

    • Coverage across research, market analysis, protocol operations, and execution support

    • Clarity around when a workflow needs setup, access, or live service context

    Fixed

    • Agent selection of the most relevant skill or access path for a task

    • Coverage across research, market analysis, protocol operations, and execution support

    • Clarity around when a workflow needs setup, access, or live service context

  • This update establishes a stronger visibility layer for agent execution. Instead of leaving users to guess what happened between intent and output, AABC now keeps more of the workflow observable across generation, accumulation, and preview states. The result is a product that feels less like a black box and more like a guided execution surface.

    Added

    • A coordinated execution visibility layer for multi-step runs

    • Better accumulation of outputs across related tasks instead of treating each step in isolation

    • Clearer preview handoff between generation and visible result states

    Improved

    • Preview continuity across longer task chains

    • Synchronization between intermediate execution and final visible output

    • User confidence during long-running generation and build flows

    Fixed

    • Cases where only the first task result appeared in preview

    • Broken continuity between sequential generation steps

    • Output state loss caused by context drift in multi-step runs