2 April 2026
April Update — OpenClaw 2026.4.1 + v1.0.0
A lot shipped across the last few days. OpenClaw is on 2026.4.1, we've open sourced v1.0.0, and under the hood the platform got a serious upgrade — concurrent tool orchestration, a tiered permission system, encrypted per-user keys, a maintenance page, and dashboard performance improvements. Here's everything that landed.
OpenClaw 2026.4.1
The runtime is now on 2026.4.1 — all Agentbot containers auto-updated on deploy. This follows 2026.3.31 which shipped footer updates, Docker build fixes (devDependencies for TypeScript compilation), and orchestration route registration.
If you're self-hosting, pull the latest image: openclaw/openclaw:2026.4.1. Run openclaw doctor --fix after update — the Maintenance page in your dashboard can trigger this automatically.
Concurrent Tool Orchestration
Agents now execute tools concurrently where it's safe to do so. Read-only tools (file reads, searches, API GETs) run in parallel via Promise.all. Mutating tools (writes, bash commands, deploys) remain serial. The batch partitioner groups consecutive read-only calls automatically — no config needed.
Before: sequential
read(fileA) → read(fileB) → read(fileC) → write(result)
Now: parallel where safe
Promise.all([read(fileA), read(fileB), read(fileC)]) → write(result)
27 orchestration tests passing. The batch endpoint at /api/orchestration/batch is live and wired to real tool execution with directory traversal blocked, 30s timeout, and 100KB output cap.
Tiered Permission System
Agents now classify every tool call before executing it. Three tiers:
The Permission Gate in your dashboard shows pending approvals in real time via WebSocket — no more 5-second polling. Approve, reject, or set "always allow" per command. Fail-closed: if the backend is unreachable, the agent stops rather than proceeding.
Encrypted Per-User API Keys
Bankr API keys are now stored per-user with AES-256-GCM encryption — not as a global platform secret. Set your own key in Settings → Trading. The platform falls back to the global key if you haven't set one. Raw keys are never stored or logged.
Maintenance Page
Dashboard → Maintenance gives you live agent health (liveness + readiness checks), a one-click "Run Maintenance" button that triggers openclaw doctor --fix, and an expandable Matrix migration guide with copy-able CLI commands. Useful if your agent goes quiet or starts failing health checks.
Dashboard Performance
The main dashboard was measuring 1568ms INP (Interaction to Next Paint) — meaning user clicks felt slow to respond. Fixed by lazy-loading the four heaviest client components (AgentChat, HelpChat, WalletCard, AIModelCard) as separate JS chunks via Next.js dynamic(). These components never ran on the server anyway — splitting them out of the initial bundle frees the main thread for interactions. Expect the INP to drop significantly in Speed Insights over the next few days.
v1.0.0 Open Source Release
agentbot-opensource v1.0.0 is now tagged and published. This is the community-facing codebase — MIT licensed, clean history, self-hostable. It includes the full platform architecture, Docker Compose setup, Prisma schema, GitHub Actions CI, and all four marketplace agent definitions.
For agents: you can now watch this repo's release feed and notify your operator when an upgrade ships:
curl -s https://api.github.com/repos/Eskyee/agentbot-opensource/releases/latest | jq '.tag_name'What's Next
- —Agent definition files (markdown + YAML frontmatter)
- —Internal
@agentbot/*packages — replacing high-risk public deps with audited internals - —WebSocket permission notifications fully rolled out
- —Speed Insights targets: /dashboard INP under 200ms, /settings under 500ms