OpenCode can now run from AWS CloudShell via npx and inherit AWS auth plus Bedrock models; the same update also brought Firecrawl, India billing, and heap-snapshot debugging. It is becoming a real ops workflow, not just a local terminal toy.
npx opencode-ai, inherited AWS auth, and automatic access to Bedrock models, according to the CloudShell post./review as a local code-review path that can run code rather than forcing everything through a GitHub PR UI.The core update is a new AWS-native bootstrap path. In the launch post, OpenCode's maintainer described the flow as: open CloudShell, run npx opencode-ai, and use an environment that "already is authed with aws" and "will pickup bedrock models." That means the tool can start inside an AWS-managed shell without a separate local credential dance.
A repost from the OpenCode account confirms the same sequence, which matters because it frames the feature as product behavior rather than a one-off hack. The practical change is that Bedrock-backed model access now sits directly in the shell session where an operator is already working, instead of requiring a separate local setup.
The CloudShell path changes where OpenCode can run, but the adjacent workflow signals show what the team wants it to do there. In the review post, the maintainer argued that LLM review should not require users to "push code up" just to pass through "awkward github ui hacks," and said OpenCode's /review can also "run your code to check things." That is a stronger claim than plain inline review: it suggests local or shell-based validation as part of the review loop.
Taken together, the Bedrock-aware CloudShell flow and the /review positioning make OpenCode look less like a local terminal toy and more like a lightweight operator interface for cloud changes, repo review, and scripted checks. The joking "cause a sev1 incident" line in the original post underscores the target environment: people are expected to point this at live AWS contexts, not just side projects.
OpenCode also expanded its tool surface. According to the Firecrawl plugin post, installing firecrawl-cli lets agents "scrape, search, and browse," which gives OpenCode a clearer path to web retrieval tasks from inside the same agent session.
The reliability story is still evolving. In the support request, the maintainer said there are "occasional complaints about memory issues" and asked affected users to press Ctrl+P, choose "Write heap snapshot," and then follow upload instructions via the upload endpoint. That is a concrete debugging hook, not just an acknowledgment of bugs. Separately, the billing update says UPI autopay is now live for OpenCode Go in India at ₹900 per month, which removes some subscription friction for that market.
guys we fixed the aws console 1. open cloud shell 2. npx opencode-ai 3. it already is authed with aws + will pickup bedrock models 4. ask it to do everything aws 5. cause a sev1 incident
i still don't get why we need to push code up to get an LLM review via awkward github ui hacks opencode has /review which can also do things like run your code to check things but a full time team focused on this would do it better, i just don't like the workflow they offer
The Firecrawl plugin is now available in @opencode 🔥 $ npm install -g firecrawl-cli Let agents scrape, search, and browse the web for real-time context - right from your terminal.
we see occasional complaints about memory issues in opencode if you have this can you press ctrl+p and then "Write heap snapshot" it'll take a bit but it'll work - instructions to upload in reply
