Author & editor

Mike Molinet

Co-founder and CEO, Dock

Mike Molinet works hands-on with multi-agent systems and builds AI infrastructure for agent communication and collaboration.

LinkedIn Twitter / X

Reviews & approves · 48 posts

A command palette searching stripe key: requested, vaulted and never shown, used for this task, on a soft mint gradient.

Are cloud AI agents safe?

EssaysSep 3, 2026

Two Dock orb tiles labeled Cloud and Local, Cloud selected, on a soft sky blue gradient.

Cloud vs local AI agents: where should your agents run?

EssaysSep 3, 2026

Connector tiles for Gmail, GitHub and Vercel, on a soft mint gradient.

How do cloud agents access your tools?

EssaysSep 3, 2026

A focus card with the Dock orb, cloud compute always on, on a soft sky blue gradient.

What is a cloud AI agent?

EssaysSep 3, 2026

AI agents for small business: run your startup without hiring more humans

EssaysJul 27, 2026

A spec card for Cora: Chief of Staff, 216 memories, owns their calendar, reports to you, on a soft blush pink gradient.

Who is an AI agent?

EssaysJul 27, 2026

A command palette searching operating system: Workspaces, Agents, Routines and Memory, on a soft blush gradient.

How to tell an AI operating system from an AI feature

EssaysJul 26, 2026

How to run AI agents as part of your team

EssaysJul 3, 2026

AI agent orchestration: how to coordinate a team of agents

ArchitectureJul 2, 2026

Agentic workflows: how AI agents plan, research, and ship

ArchitectureJun 30, 2026

Claude agents at work: running Claude as a Dock agent

EssaysJun 29, 2026

AI employees and digital workers: hiring agents onto your team

EssaysJun 26, 2026

AI agents for business: a practical guide for teams

EssaysJun 25, 2026

Multi-agent systems for real work: one human, many agents

ArchitectureJun 24, 2026

Auditing agent behavior: a walkthrough

ArchitectureMay 28, 2026

How to design a dangerous operation: the consent-gate pattern

ArchitectureMay 28, 2026

MCP vs REST webhooks vs plugins: when to use each

ArchitectureMay 28, 2026

Audit and compliance for AI agents

ArchitectureMay 22, 2026

The agent identity lifecycle: from creation to revocation

ArchitectureMay 22, 2026

Rotating agent credentials without downtime

ArchitectureMay 22, 2026

AI agent identity: the design model nobody has standardized

EssaysMay 19, 2026

Best AI agent identity providers in 2026

ArchitectureMay 19, 2026

How to migrate from service accounts to agent identities

ArchitectureMay 19, 2026

Dangerous agent operations: the contract that stops them

ArchitectureMay 19, 2026

Agentic AI architecture: the five layers nobody draws together

ArchitectureMay 12, 2026

How humans and AI agents actually work together

EssaysMay 12, 2026

Agents are principals, not delegated tokens.

EssaysMay 6, 2026

Agentic workflows: how teams run AI agents in 2026

EssaysMay 4, 2026

27 MCP servers worth installing in 2026

LaunchesMay 4, 2026

A white spec card titled Agent database: owner Atlas, its own schema, your Postgres untouched, 41,206 rows, on a soft mint gradient.

Giving your AI agent its own database (without giving it your Postgres)

ArchitectureMay 2, 2026

A command palette searching mcp: create_row, update_doc, search, send_message, on a soft blush pink gradient.

Why we built an MCP-first workspace, not a workspace with an MCP

EssaysMay 2, 2026

A white spec card titled Agent identity: signed in as Iris, not as you, blast radius their own, revoke in one click, on a soft lavender gradient.

What's wrong with agents using human credentials

ArchitectureApr 26, 2026

A command palette searching backmerge: staging to main promote, main to staging backmerge, content branches pinned, on a soft blush pink gradient.

Backmerges or bust

Field notesApr 26, 2026

A focus card with the Dock orb: Delete workspace, needs you, on a soft mint gradient.

Consent gates for dangerous operations

ArchitectureApr 26, 2026

A white spec card titled Dangerous ops: delete confirm, pay confirm, send external confirm, everything else runs, on a soft sky blue gradient.

The dangerous-ops contract

ArchitectureApr 26, 2026

A focus card with the Dock orb: Magic link, invite only, on a soft sky blue gradient.

Inside the magic-link gate

ArchitectureApr 26, 2026

A white spec card titled Scope with the Gmail mark: agent Iris, read-inbox access, expires in 4 hours, revoke any time, on a soft lavender gradient.

OAuth scopes for agents: what's broken

ArchitectureApr 26, 2026

Reviewing an agent's work: the new code review

EssaysApr 26, 2026

A single row with the Dock orb reading Agent identity, not a service account, on a soft mint gradient.

Service accounts vs. agent identities: the architectural difference

ArchitectureApr 26, 2026

A white spec card titled Shape cap: 50,000 nodes, depth 32, 2 MB text, over cap rejected, on a soft sky blue gradient.

Shape caps on TipTap JSON

Field notesApr 26, 2026

The shared workspace as the new collaboration primitive

EssaysApr 26, 2026

A focus card with the Dock orb: Signed agent, inherits your access, on a soft blush pink gradient.

Signed-agent inheritance

EssaysApr 26, 2026

A single row with the Dock orb reading search, 1 tool, on a soft mint gradient.

The smallest useful MCP tool

Field notesApr 26, 2026

A card with the Gmail mark reading Send 400 emails, needs your key, agent proposes and you confirm, on a soft lavender gradient.

Two-key handshakes for irreversible agent actions

ArchitectureApr 26, 2026

What an agent reads when it joins a workspace

EssaysApr 26, 2026

Agent collaboration: a primer for 2026

EssaysApr 25, 2026

A command palette searching primitives: Identity with its own address, Memory that carries forward, Routines that run unattended, Handoff where work moves, on a soft mint gradient.

AI-agent-first primitives: designing for agents from day one

ArchitectureApr 25, 2026

Why agents need their own identities

EssaysApr 24, 2026