Directory AI agents Autonomous SWE agent

Amp: AI agent (Autonomous SWE agent) in 2026

Coding agent with remote 'orbs' where agents keep working in the cloud.

Sourced facts checked 2026-10-17 Rank 3 of 6 in Autonomous SWE agent. Source: Official site or docs, checked 2026-10-17.

Key facts

Type	AI agent
Category	Autonomous SWE agent
Maker	Amp (formerly Sourcegraph)
Interface	CLI, IDE, web
MCP support	Yes (per official docs)
Subagents	Yes
Headless mode	Yes
Pricing (official)	Hobby free (pay-as-you-go, BYOK); Individual $20/mo; Teams no extra charge; Enterprise custom. (checked 2026-10-16)
Note	Maker attribution not re-checked.
Website	https://ampcode.com
Source	https://ampcode.com (checked 2026-10-17)
Vibe Shipped harness page	https://vibeshipped.io/harness/amp
Vibe Shipped toolbox page	https://vibeshipped.io/agent-toolbox/amp

Install

This agent is an MCP client (official docs). To give it a server from this directory, add that server in this agent's own MCP settings; every MCP server entry shows the snippet.

Source: https://ampcode.com (checked 2026-10-17).

Is this your ai agent? Claim this listing to correct facts, add a contact, or tell us you want the upcoming verified badge, downtime alerts and agent-demand analytics. Claiming never changes where it ranks.

Other autonomous swe agent ai agents

OpenHands Sourced facts checked 2026-10-17 · #1 in Autonomous SWE agent
Open-source autonomous SWE agent with web GUI, CLI and SDK.
Roo Code Sourced facts checked 2026-10-17 · #2 in Autonomous SWE agent
Open-source VS Code agent with role-based modes; makers now focus on Roomote cloud agent.
Jules Sourced facts checked 2026-10-16 · #4 in Autonomous SWE agent
Google's asynchronous coding agent: clones your repo to a cloud VM, plans, and opens a PR.
Devin Sourced facts checked null · #5 in Autonomous SWE agent
Cognition's autonomous AI software engineer that takes tickets and opens PRs.
Factory (Droids) Sourced facts checked 2026-10-16 · #6 in Autonomous SWE agent
Agent-native development platform: 'Droids' run in desktop, CLI, SDK and cloud background agents.

Within a category, entries sort by (1) latest verification: verified working, then reachable but sign-in required, then failing; (2) share of daily checks that passed; (3) number of sourced facts attached to the entry; (4) median response time of passing checks; (5) name. Nothing else counts. Rankings are never for sale: sponsors and paid plans cannot change an entry's position. Markdown · JSON