Product
EnterprisePricingCompanyBlogCommunityDocsD
Back

March 6, 2026 Update

Product UpdatesDeveloper ExperienceCoderDeployAnalytics

Since the last update, we shipped 8 releases and 47 commits across the SDK. This week: Agentuity Coder, one-click GitHub deployments, new CLI flags, query-level database analytics, and Agent Skills. Let's get into it!

Overview

  • Agentuity Coder with remote sandbox sessions and task chains
  • Deploy Now for one-click GitHub project deployment
  • New CLI flags for structured input, schema introspection, and output filtering
  • Database query analytics with latency percentiles, slow query detection, and interactive schema diagrams
  • Agent Skills that teach AI coding agents about the Agentuity SDK and patterns
  • And more!

Agentuity Coder

The new @agentuity/coder package is a dedicated terminal interface for the Coder Hub. It connects your local terminal to remote sandbox sessions so you can launch, monitor, and orchestrate coding agents from one place.

# Start a coding agent session
agentuity coder start

# List active sessions
agentuity coder ls

# Inspect a specific session
agentuity coder inspect codesess_abc123

A few key features:

  • Remote sessions. Attach to a running sandbox from your terminal. Pick a session interactively or connect directly with a session ID. The CLI polls for readiness (up to 120 seconds) before attaching, so you won't connect to a session that's still booting.

  • Chain editor. A keyboard-driven overlay for composing sequential and parallel agent task chains. Build multi-step workflows visually and send them for execution.

  • Live monitoring. Sessions, tasks, and participant counts stream in real time with automatic reconnection if the connection drops.

If you've been following the OpenCode plugin and session dashboard additions, Coder is the next evolution: a purpose-built client that brings all of these together in one terminal experience. This package replaces the previous @agentuity/pi.

Deploy Now

Deploy a GitHub repository to Agentuity in one step:

agentuity project import --source github --remote owner/repo

The CLI validates your GitHub identity, provisions any required resources, and streams build output in real time. Environment variables are parsed from the project's agentuity.json template, with required/optional labels and descriptions shown inline.

The web console has a matching Deploy screen that shows the same streaming output, so you can follow along from either the terminal or the browser.

New CLI Flags

Four new global flags make the CLI easier for AI agents (and scripts) to work with:

--input accepts arguments and options as a single JSON object:

agentuity cloud sandbox create --input '{"runtime":"bun:1","memory":"1Gi","network":true,"port":8080}'

--describe returns the full schema for any command without executing it. No authentication required:

agentuity cloud sandbox create --describe
# Returns: { "name": "create", "options": [...], "arguments": [...] }

--fields filters JSON output to just the fields you need:

agentuity --json --fields "id,name,status" cloud deployment list

--validate checks your input against the command schema without running anything:

agentuity --validate cloud kv set --input '{"namespace":"ns","key":"k","value":"v"}'
# Returns: { "valid": true, "command": "cloud kv set" }

Together with the existing --json and --dry-run flags, these enable a discover-validate-execute workflow: introspect the schema, validate the input, then execute with structured I/O.

Database Query Analytics

A new Performance tab on the database detail screen gives you query-level observability. This extends the service analytics we shipped last week with database-specific depth.

At the top, summary cards show P50, P95, and P99 latency alongside total query count and error rate. Below that, two charts plot latency percentile trends and query volume over time with a configurable time range (last 24 hours, 7 days, 30 days, etc.). A Slow Queries table ranks queries by total time impact, showing the query pattern, command type, call count, average duration, P95 duration, total time, average rows, and error count.

We also added an interactive schema diagram to the Database Schema tab. Powered by React Flow, it renders tables as draggable nodes with columns and draws animated edges for foreign key relationships. Toggle between the visual diagram and the existing table view.

Agent Skills

AI coding agents, including the ones from our agent onboarding guide, can now learn Agentuity SDK patterns on the fly:

npx skills add agentuity/sdk

This installs four skill modules that follow the Agent Skills specification:

  • agentuity-agents — Building agents with createAgent, schemas, context APIs, state management
  • agentuity-cli — Project scaffolding, dev server, deployment, cloud services
  • agentuity-routing — API routes, middleware, WebSocket/SSE/WebRTC handlers
  • agentuity-workbench — Interactive dev UI for testing agents

Each module gives the coding agent enough context to scaffold projects, use ctx APIs, set up middleware, and test with the dev UI. You can install them for Claude Code, Pi, Cursor, or any agent that supports the spec.

Also Shipping

SDK and Runtime

  • KV TTL extended to 365 days — TTL=0 now means no expiration. Fixed --ttl and --confirm flag parsing in the CLI. See the KV docs for the full API.
  • Queue provisioning API — ctx.queue.createQueue() and ctx.queue.deleteQueue() for programmatic queue management
  • Bucket configuration SDK — New client and CLI command for storage bucket settings with template flow prompts
  • Email IMAP/POP3 config — Connection settings for external email clients via ctx.email
  • Webhook analytics — getWebhookOrgAnalytics() and getWebhookOrgTimeSeries() for metrics and time-bucketed data
  • WebSocket improvements — Better reconnection handling with max length validation and terminal close code detection (4000-4999)
  • Biome migration — The SDK monorepo replaced Prettier + ESLint with Biome for linting and formatting
  • Task enhancements — --tag flag on task create, delete/cleanup command, and UserEntityRef type for user/agent discrimination

CLI Improvements

  • Cloud region list — agentuity cloud region list shows available regions with your default highlighted. Supports --json.
  • JSON-formatted errors — The --json flag now outputs structured error responses instead of plain text
  • Sandbox exec improvements — Separate stdout/stderr in JSON responses and directoriesCreated tracking for copy operations
  • Faster dev startup — --experimental-no-bundle and --no-typecheck flags skip bundling and type checking during agentuity dev
  • Automatic TLS provisioning — DNS validation for custom domains now triggers Let's Encrypt provisioning automatically, removing the previous manual TLS gate

Web Console

  • Webhook UI redesign — Activity charts with received/delivered breakdown, tabbed filters, 30-second auto-refresh
  • Queue analytics — Org-level activity charts, consumers tab, and destination metadata
  • New service screens — Email, Task, and Schedule each have dedicated screens with stats cards in the sidebar
  • Email improvements — Activity charts replace stats cards, per-address counts, and a new IMAP/POP3 connection settings dialog
  • Bucket settings UI — Manage storage bucket configuration directly from the console
  • Copy as Markdown — One-click export of task details

Bug Fixes

  • Auth schema: missing apikey and invitation columns added with compile-time type assertions
  • Better Auth 1.5.0 compatibility restored after breaking upstream release
  • Base64-encoded values in kv.search() results now decode correctly
  • .ts import extensions enforced for Node.js ESM compatibility
  • Queue schema improvements and API module refactoring

Examples Update

We mentioned last time that the examples repository was getting a major overhaul. That work has landed: 50+ ready-to-run projects covering a guided training path, framework integrations (OpenAI Agents SDK, LangChain, Mastra), and drop-in patterns for existing apps (TanStack Start, Turborepo). Sandbox-powered examples like a web explorer and code runner are coming soon!

Upgrade Now

The CLI will prompt you automatically when a new version is available, but you can also upgrade manually:

agentuity upgrade

Or if you're new to Agentuity, welcome! Get up and running in seconds:

curl -fsSL https://agentuity.sh | sh

Once you're upgraded, try some of the new features:

  • Launch a coding session with agentuity coder start
  • Install Agent Skills with npx skills add agentuity/sdk
  • Deploy a repo with agentuity project import --source github --remote owner/repo
  • Check database query performance in the web console

Resources

  • Documentation
  • Web Console
  • GitHub SDK
  • Full Changelog

Questions or feedback? Drop by our Discord to chat. Happy building!

Table of Contents

  • Overview
  • Agentuity Coder
  • Deploy Now
  • New CLI Flags
  • Database Query Analytics
  • Agent Skills
  • Also Shipping
  • Examples Update
  • Upgrade Now
  • Resources

The full-stack platform
for AI agents

Copyright © 2026 Agentuity, Inc.

  • Contact
  • Privacy
  • Terms
  • Security
  • Features
  • AI Gateway
  • APIs
  • Custom Domains
  • Evals
  • Instant I/O
  •  
  • React Frontend
  • Sandboxes
  • Storage
  • Workbench
  • Company
  • Enterprise
  • Pricing
  • Blog
  • About Us
  • Careers
  • FAQ
  • Links
  • App
  • Docs
  • Discord
XLinkedInYouTubeGitHubDiscord

Copyright © 2026 Agentuity, Inc.

  • Contact
  • Privacy
  • Terms
  • Security

Thought Leadership, Developer Ready (TLDR)

AI Agent InfrastructureAI Agent DeploymentAI Agent ObservabilityAI Agent RuntimeMulti-Agent Orchestration