LoginSign Up
← Back to Collab
Weekly · Collab · Week of June 15, 2026

Collaboration tools push agents from assistant to actor — writing code, running CI, and capturing meetings — with metering close behind.

agentic-workflowsmcpai-meteringgovernanceintranet-marketingdeveloper-tools
Y
By Yahya Tür
LinkedIn ↗
Generated 5h agoDrawn from 13 products

The week in collaboration

The directional move this week is agents crossing from assistant to actor inside collaboration tools. GitHub put coding agents directly into GitHub Actions with Agentic Workflows, Linear let its agent write code and open PRs without leaving the tracker, and Miro wired its canvas to coding agents over MCP. The same arc shows up at the conversation-capture layer: Claap and Avoma both opened their meeting data to external AI clients via MCP, treating their transcripts as something agents should query rather than data locked in their own UI. Across very different products, the shared bet is that the workspace is becoming a place where work is executed, not just coordinated.

The second, quieter pattern is governance and metering catching up to the AI layer. GitHub surfaced AI Credits in standard usage reports and shipped security validation for third-party coding agents; Asana now shows builders when an AI Studio rule will consume credits. In the intranet corner, much of the activity is positioning rather than shipping — several vendors (Simpplr, Interact, Claromentis, Powell, Happeo, Staffbase) lean on AI-governance essays and comparison content, with only the occasional real release underneath. The week's genuine product signal concentrates in developer-adjacent collaboration and revenue-intelligence tools; classic intranet feeds were mostly marketing.

Leaders

  • GitHub anchored the week, moving Agentic Workflows into public preview so coding agents can handle issue triage, CI-failure analysis, and doc updates from inside GitHub Actions. It paired that with GA of security validation for third-party agents like Claude and Codex, and made Claude Fable 5 generally available in Copilot — three sparks that together formalize external agents as first-class repo actors that need guardrails.
  • Mattermost turned a run of bilateral partnerships into one integrated stack, combining archTIS policy-based access, Arqit post-quantum crypto, and Whitespace into a sovereign command-and-control surface for defence ministries and coalition partners. With Agents V2 just behind it, Mattermost is converging its partnership stack and agent layer into a single regulated-collaboration pitch.
  • Linear closed its plan-write-review-ship loop: after teaching its agent to read code (Code Intelligence) and review it (Diffs), Coding sessions now let the agent write code with Claude Code and Codex and open a PR from an issue or a Slack thread. The project tracker is becoming the place work is executed, not just tracked.
  • Claap completed its mobile rollout with iOS joining Android, extending capture from virtual calls to in-person meetings, and opened its smart tables and AI columns to Claude and any MCP client while deepening HubSpot enrichment. Both moves push Claap toward being a capture layer and a data source for the wider agent ecosystem.
  • Asana had no sparks but shipped the week's most coherent improvement run: AI Studio now flags when an automation rule consumes credits, alongside RBAC for create permissions and deeper HubSpot and Slack context. The credit-awareness change is small but telling — Asana's AI automation is now metered and budgeted, not just tried.

Wildcards

  • Trilium Notes is the clear off-pattern move: while nearly every other product wires agents in, Trilium's 0.102.0 removes its built-in LLM integration outright, citing maintenance complexity, and ships it as a non-reversible database migration. In a week defined by AI integration, a deliberate AI removal stands out.

Themes that compounded

  • Agents became actors, not assistants: GitHub, Linear, and Miro all moved agents into execution surfaces (Actions, PRs, the canvas) rather than leaving them as side-panel helpers.
  • MCP is the connective tissue: Claap, Avoma, and Miro all used MCP to either expose their data to external agents or pull agent output in.
  • AI consumption is being metered: GitHub's AI Credits in usage reports and Asana's credit-awareness for rule builders both treat AI as a billable line item customers must budget.
  • Governance guardrails accompanied the agent push: GitHub's security validation for third-party agents and Mattermost's policy-enforced sovereign stack show controls shipping alongside capability.
  • Intranet vendors mostly published, not shipped: Simpplr, Interact, Claromentis, Powell, Happeo, and Staffbase feeds were dominated by AI-governance thought-leadership and comparison content rather than product changes.

Watch this week

The agent-as-actor and MCP threads are the ones with real momentum in this feed, so watch whether the metering and governance work keeps pace with the capability. GitHub already ties AI Credits into billing and validates third-party agents; Asana is surfacing credit consumption in the build loop — the open question is how far the permission, approval, and cost controls extend as agents gain write access. On the intranet side, the gap between AI-governance messaging and shipped governance product (Simpplr's AI Control Center being the rare concrete launch) is worth watching: several vendors are selling oversight while their feeds show mostly essays, not releases.