MeshMeshBeta
Releases

October 16, 2026 - Deep scans you can act on

Map systems now runs deep, read-only scans of your org that end in one clear verdict and ranked moves you can plan with an agent. Plus redesigned connections, and new agents that start with your workspace's shared context.

October 16, 2026

Desktop 0.21.15

Knowing what's in your org is the first step to changing it. Map systems now runs deep scans with a team of specialist agents, and every scan ends in a live report that says what matters, why, and what to do next, with a button to plan each move with an agent.

What's New

Deep scans in Map systems

Start a scan from Map systems → New analysis. Pick a focus (Full org health, Security & access, Automation, Data quality, Code quality, Data model, Pipeline or Custom), check the brief, choose the org it reads, and start.

  • Six specialist agents map your org in parallel: adoption and value, automation and code, data model and governance, security and access, apps and packages, and business context
  • Watch it work: a live map of your org, a run log with each step, and the agents as they spawn
  • Scans run in Research mode: they read your connected systems but can't change them
  • Every analysis gives you a live report and a business context your agents start from
  • Your analyses are private to you, and listed in one place under Your analyses, next to the workspace report

Org scans you can act on

Every scan now ends in one live report canvas instead of a static document.

  • One clear verdict: a short posture and a one-line thesis about your org, with the key numbers underneath
  • The moves that matter most, ranked by leverage: each move says what to do, why (with the real numbers behind it), how, the effort, the impact and who owns it
  • Plan with an agent: start an agent from any move. It starts in Plan mode with the report attached, so nothing changes until you approve the plan
  • Recommendations on your dashboard show their progress: an agent is working, needs you, or has run

Connections, redesigned

Connecting Salesforce and your other tools is quicker and clearer.

  • Salesforce comes first, with a Start here hint
  • New Salesforce connections connect to a Sandbox by default
  • A new Salesforce connection is named after its org, so you can tell your orgs apart
  • On the web, tools like Linear, Notion and Atlassian sign in through a popup that closes itself when you're done, and your connection lists update straight away
  • Switch a connection off or back on from its details with Disable connection and Enable connection

Agents start with your context

Your workspace's org report, business context and brand kit are now default references. Every new agent starts with them attached, so it knows your org and your brand from the first message.

  • They show as pills in the composer. Remove any of them for one agent
  • Admins choose them in Settings → Default references
  • Agents started from Slack or over MCP get them too
  • Brand kits have a new view: your workspace's default kit up front, every other kit below, and Make default for admins
  • A canvas looks the same everywhere: it uses the brand it was made with, otherwise your workspace's default brand kit

What's Improved

  • Guided setup for new workspaces — the first admin goes from brand kit to connecting Salesforce to mapping your systems in one flow, and invited members get a short tour of agents, workbenches, the Library and skills
  • Discover — start from public and workspace recipes, skills and collections, right from your dashboard
  • A refreshed sign-in page — new artwork and the MeshMesh from Salesforce lockup
  • A faster agent list

What's Fixed

  • Agents that finished no longer read as "Analysing…" forever
  • Live agents are no longer shown as paused by mistake
  • A new brand kit or scan no longer replaces your workspace default

Security & reliability

  • Org scans and business-context agents can't change your connected systems
  • Agents can only get access to the connections they were given
  • Stronger checks on agent file paths and storage paths
  • Dependency security patches and behind-the-scenes hardening