FellowMakers.app
‹ All wavesAgents, context & control
Wave11 products

Constrain and inspect agent actions

People need explicit limits or inspectable records around what an agent does.

Different approaches

Dedicated disposable machines isolate execution. Approval gates, policy checks, scoped tools and spending caps restrict actions. Repository contracts, issue-linked diagnoses and visual action displays expose what ran and what it produced; these mechanisms operate at different layers.

What brings them together
An explicit isolation boundary, action approval or authority check, spending cap, execution-proof contract or inspectable agent-action trail.
Outside this wave
Agent hosting alone, cloud tools without stated controls, messaging, shared knowledge, scaffolding and generic incident automation.
How the waves were researched
Where it sits on the board
AI infrastructureDeveloper toolsAIProductivitySecurity
Builders in Europe 4 · North America 1 · Asia 1 · 5 not placed
In the wave11
Amakoraamakoragroup.com · AI infrastructure · by Tochii

Enterprise infrastructure for governed AI execution.

“Building the infrastructure between AI models and real-world execution. Founder, https://t.co/bAFkVXhXro. AI products, agents and developer platforms. Stanford GSB.”Collieheycollie.com · AI · by Rick A.F.

Open-source desktop AI assistant that executes approved task plans.

“You describe the job in plain English. Collie shows the plan, you approve, it gets done.”Foveawww.hellofovea.com · Productivity · by Max

Mac interface for coordinating work across AI agents.

“Fovea knows your next step. The last layer between you and every agent you run. Get Early Access Fovea for Mac is currently in Alpha. Request Access. Nothing to attach.”Mcpifexmcpifex.com · AI infrastructure · by AJawhar

Marketplace and gateway for MCP servers.

“Issue scoped keys per agent”NoBurnnoburn.dev · AI infrastructure · by Robat Das Orvi

Pre-request spending controls for LLM APIs.

Otaota.run · Developer tools · by bobai

Repository execution-contract tooling for people, CI and AI agents.

Polyusepoly.co · Developer tools · by mitchell

Shared workspace for teams building with AI.

“Poly is a shared room where you and your team build with AI — together. Everyone sees every change as it happens, and anyone can approve it, undo it, or take the wheel.”Sandbox as a Servicesandbox-as-a-service.com · AI infrastructure · by Florian S

Dedicated virtual-machine sandboxes for agents.

“A minute later your agent has a dedicated virtual machine it can break — and you have not installed an SDK, configured a cluster or spoken to anyone.”SentinelSCAsentinelsca.com · Security · by SentinelSCA

Execution-governance layer for consequential AI-agent actions.

“Sentinel checks identity, authority and policy before a consequential action executes.”Termitermiprotocol.com · Developer tools · by Omer Ercan

Live 3D observation workspace for AI coding agents.

Thermitethermite.rs · Developer tools · by Hauke 🌱🦀

Self-hosted error tracker that hands issues to a user's coding agent.

Related waves
When the coding agent becomes a teammateMcpifex, NoBurn, Ota, Sandbox as a Service, SentinelSCA, Termi, Thermite7Give agents a working interface to existing systemsMcpifex, Thermite2MCP is the new APIMcpifex, Sandbox as a Service2Prepare work for human reviewCollie1