Custom coding agents with skills and MCP tools

Aiola saves coding agents with instructions, memory and skills, and keeps MCP servers per workspace, in one macOS and Windows desktop app. Each repeat job starts with the context and tools it needs.

Last updated November 12, 2026

Key facts

Give agents saved instructions, skills and memory so repeat tasks start with useful context.
Keep each workspace's tool connections separate and inspect the tools a server offers.
Match skills to a compatible provider and review changes before trusting a repeated job.

Save the agent's job and context

Create an agent inside the project workspace with instructions, memory and linked skills. Templates provide starting points. Review those copied rules before saving your profile.

Attach reusable instructions

Inspect skill files before attaching them to an agent or starting a chat. The first skill locks the compatible provider. Edit workspace action prompts for recurring requests too.

Choose tools for this workspace

Add stdio or HTTP MCP servers in Workspace Settings and enable the ones you need. Expand each row to discover tools. These connections belong to the current project workspace.

Follow the workflow

  1. Create a workspace agent with a clear role, instructions and memory.
  2. Inspect a skill's files and attach skills compatible with the same provider.
  3. Adjust analysis and fix prompts under Workspace Settings, Workspace, Instructions.
  4. Add an MCP server, enable it and expand its row to discover tools.
  5. Choose the saved agent for a thread or automation and inspect its first result.
  6. Use local profiles and third-party servers; hosted reports and hosted MCP are part of Pro.

What to review at the end

  • A saved agent whose instructions and memory belong to one workspace.
  • Reusable skills matched to the provider that can load them.
  • Project-specific MCP connections with tools you inspected before use.

See all workflows and guides.

Common questions

Everything your agents need, in one app

Errors, user feedback, GitHub and analytics live next to your coding agents. Every issue becomes a task an agent can take.