Coder Makes Its Self-Hosted AI Agents Generally Available
Software / news
Coder Makes Its Self-Hosted AI Agents Generally Available
Nearly 70% of beta usage came through the API rather than a chat window, and the round funding the launch closed five months earlier at $90 million.

Coder took Coder Agents out of beta on Sept. 9, 2026, the company said in a blog post, pitching it as a way for regulated enterprises to run AI coding agents entirely on infrastructure they control instead of a vendor's hosted control plane.
The Austin, Texas-based company's post, written by product marketing manager Matt Vollmer, says both the agent and its execution environment now stay on the customer's own infrastructure, including air-gapped networks with no outside connection at all. That is the pitch aimed at government and financial-services customers who cannot send code to a third-party agent host.
Coder was co-founded by Ammar Bandukwala and Kyle Carberry, according to the company's own about page, which says its open-source tooling now has a community of more than 100,000 developers, separate from the enterprise customers Coder Agents targets.
What Coder Agents GA actually changes
General availability adds stable APIs for headless, background agent work, deeper role-based access control and auditing, and controls over which MCP servers and models an agent can reach. MCP, the Model Context Protocol, is the open standard agents use to call external tools, the same standard behind the skill marketplaces now drawing security scrutiny. Coder pairs that access control with its AI Gateway, which the company says gives platform teams observability into model usage and a way to cap spending per team, joining other developer tools tightening session controls this month, including Warpgate's 0.29.0 release.
How enterprises actually used the beta
The most specific number in Coder's post is about behavior, not features: nearly 70 percent of Coder Agents workloads during the beta were invoked through the API rather than through interactive chat, Coder said. That points to enterprises wiring agents into CI pipelines and scheduled jobs more than using them as a chat-based pair programmer.
| Metric | Figure | As of |
|---|---|---|
| Series C round | $90 million, led by KKR | Apr. 6, 2026 |
| Year-over-year bookings growth | 300% | Apr. 6, 2026 |
| Beta workloads invoked via API | About 70% | Sept. 9, 2026 |
What the $90 million round is paying for
Coder closed that Series C on April 6, 2026, led by KKR with Qube Research & Technologies also participating, Built In Austin reported. Chief Executive Rob Whiteley said at the time: "We chose to partner with customers KKR and QRT in this round because they have the clearest view of how AI evolves software development." Coder told Built In Austin its bookings grew 300 percent year over year across the four quarters before the round closed, and said it planned to use the money to expand into Europe, Asia and North America.
What Coder has not said
Neither the GA post nor the funding coverage names a single customer running Coder Agents in production, and Coder has not published what the AI Premium tier costs on top of the open-source Community edition. The 70 percent API figure comes from Coder's own telemetry during its beta, with no outside audit cited, so it describes what beta customers did, not what every enterprise buyer will do once the meter is running for real.
Sources
More in Software
- 01TIN Beats ParadeDB by 57x in PlanetScale's Own Postgres TestThe extension is listed as a supported community extension on PlanetScale's own Postgres product, but there is no independent repository, published license, or way to install it outside that service.
- 02SearXNG's Creator Ships Hister 0.19, a Search Engine for OneAdam Tauber's self-hosted index now extracts full ChatGPT and Hacker News threads and speaks the Model Context Protocol, so an AI agent can search a user's own browsing history instead of the open web.
- 03Cua Ships CUA-S1-Forms With a Checkpoint It Can't LoadThe 706,048-parameter model scored 99.7 percent against a rival's hosted service in Cua's own test, but the file it shipped needed a same-day fix before its own code would open it.
- 04Addy Osmani's Skills Repo Passes 97,000 Stars at AnthropicOsmani built the 25-skill collection while still at Google, and its listing carries a passing Snyk scan, the same firm whose February audit found flaws in more than a third of the skills on the two marketplaces where his repository is also listed.