Amika (YC Fall 2025): teardown vs XMarts

In one line: Amika rents out cloud computers ("devboxes" or "Rigs") where coding agents such as Claude Code and Codex work, for engineering teams, and can run them on the team's own machines for free [site] [docs]. Threat: 2 because it is the same buyer as W1 (small engineering teams running coding agents), live and priced, but it is two people, has no named customers, and sells the place agents run, not trust in the result [site] [yc].

1. Fact file

2. Their proposition, reverse engineered

3. Positioning

• Category they claim: Cloud devboxes for coding agents; earlier, an "operating system" for coding agents and a "software factory" [site] [yc].
• Category a YC partner would put them in: developer infrastructure for coding agents (sandboxes and environments). Direct rivals in that space sell the same box by the hour (inference; the comparison page was not readable).
• What they want to be known for: the place where your agents and your team work together, on machines you already own.
• Where they sit on the map: layers L1/L2/L3 = 1/2/1, buyer = sme, developer.

4. Three layers vs ours

5. Same and different

6. Wedge overlap

7. Whitespace (where they stop)

• Layer gap: they host the agents and name checks and rules, but their public docs describe no learning from checked results [docs]. To own it we need a rule-change log tied to checked runs, with rework falling over several weeks on a real team.
• Number gap: no named customer and no result, only "we use it ourselves" [yc]. To own it we need one team's workflow with a before/after number (for example pull requests merged a week, or hours until a bug fix).
• Buyer gap: not true. They already target the small team we want [site].
• Price gap: not true. $30 a user and free on your own machines is cheap for a small team [site].
• Speed gap: not true. Free tier and self-serve start in minutes (inference from the plan page) [site].

8. For the YC application

• Proven today: Marton's logs show his system has run about 5,600 jobs with agents from three makers, each checked by a different AI and then tests, and about 160 failure records each added an automatic check. Amika's public pages show hosting and rules, but no customer and no result [site] [yc].
• Draft answer to "how are you different from Amika?": DRAFT. "Amika gives your agents a computer. We give you checked work you can trust, and every failure becomes a rule. Our system has run that loop on our own business for five months." Founders say it their own way.
• Future claim (needs proof): "A small team's agents ship checked code overnight, and the share that needs rework drops each week." Needs one outside team run for 4+ weeks, with a before/after number, and a runner that is not tied to Marton's Macs.
• What we understand that they don't: the hard part is trusting the output, not the box it runs in, and trust comes from checks that feed back into rules. proven only on Marton's own work (offer feedback lines 44 to 18); needs proof on a customer. Be careful: their launch post already names verification and guardrails [yc].
• Copyability (M3): medium. They already list verification, guardrails and audit trails as primitives [yc], so a learning loop fits what they say they will build. Against that, their current pages and docs lead with devboxes and say little about checks or learning [site] [docs], the team is two people, and their money comes from compute and seats, not from owning the result (inference). Not hard, so M3 can not score above 3 against Amika alone.
• Why now (M4): None found in this teardown. Do not use "coding agents are good now": that is general AI progress.
• Do not claim: that nobody checks agent code (their launch post says they do) [yc]. That we are cheaper or faster to start (they have a free tier). That Amika has no customers (UNVERIFIED: absence of named logos is not absence of users). Any merge rate or reliability claim for our system: only about 15% of jobs reached the main code (system inventory).

9. Ideas worth taking

1. Run on the buyer's own machines for free and charge for the team layer. It removes the security worry and the cost worry in one move.
2. Charge per active user, not per seat, so nobody has to ask for a licence to try it.
3. Ship the core as open source and sell the cloud and control layer. It builds trust with developers fast.

10. Unknowns

• Whether they have paying customers. Look for case studies on the docs site, their Discord, and the next launch post.
• What the "vs. Others" comparison page says and which rivals they name. The page text was not readable by fetch; open it in a browser.
• Real funding. Check the YC page and any seed announcement.
• Whether "automated verification" is shipped. Read the rest of docs.amika.dev (security, workflows) and try the free tier.
• Whether they sell in the UK. No UK sign found; check for a Europe cloud region.

Sources

• amika.dev [site]
• Amika pricing [site]
• Amika comparison page [site] (text not readable)
• Amika docs [docs]
• gofixpoint/amika on GitHub [docs]
• Amika releases on GitHub [docs]
• YC company page [yc]
• YC jobs page for Amika [jobs]
• YC launch post: Amika, the operating system for AI agents to write and verify code [yc]
• Amika lets coding agents keep their VMs between pull requests, 20 Sep 2026 [press]
• Dealroom listing [press] (funding figure UNVERIFIED)