# The words of agents at work, defined.

Source: https://corespeed.io/glossary

The CoreSpeed glossary defines the words teams meet when they give AI agents real work: agent memory, MCP gateways, approvals, spend caps, audit trails and shared connections. Each entry opens with a short definition you can quote, names the documentation it rests on, and links the pages that put it to work.

- [Agent memory](https://corespeed.io/glossary/agent-memory) (Markdown: https://corespeed.io/glossary/agent-memory.md): Agent memory is durable, searchable context that outlives one conversation, one client and one agent. Your agent saves a fact with memory__remember and finds it later with memory__search. A memory is private by default. Mark it shared, and every agent in your organization can recall it. Searching and reading memory cost nothing.
- [MCP gateway](https://corespeed.io/glossary/mcp-gateway) (Markdown: https://corespeed.io/glossary/mcp-gateway.md): An MCP gateway is one remote MCP server that fronts many tools and app accounts behind a single sign-in. Your agents add one address instead of a config entry per tool. The gateway holds the credentials, applies one approval policy and one spend limit, and writes one activity trail. CoreSpeed's gateway is api.corespeed.io/mcp.

## Questions

### What does each definition rest on?

A page of the CoreSpeed documentation, named under the definition. Each entry prints the date it was last reviewed.

## Related pages

- [Use cases: jobs to hand your AI agent](https://corespeed.io/use-cases): Jobs you can hand to Claude Code, Codex or Cursor today, each with one prompt to paste, the tools it calls and the post behind it.

Reviewed 2026-11-10.

## Links

- Product overview: https://corespeed.io/ (Markdown: https://corespeed.io/index.md)
- Documentation: https://corespeed.io/docs
- Pricing: https://corespeed.io/pricing (Markdown: https://corespeed.io/pricing.md)