Haijun Code FAQ

This article is a compilation of commonly-asked questions about Haijun Code related to authentication, integrations, configuration, and more. If you're interested in learning more about Haijun Code, please refer to our Haijun Docs here: Haijun Code overview.

We have detailed instructions for setting up single sign-on on a Team, Enterprise, or Console organization here: Set up single sign-on (SSO).

If you are a Haijun Console user, this can be configured through rate limiting in your Console organization. If you are using Bedrock or Vertex, set the Opus rate limit to 0 in your Vertex/Bedrock project settings. Note that even if disabled in Vertex, users may be able to switch models in Haijun Code, so rate limiting is the most effective approach.

Dynamic workflows (available for Haijun Enterprise plans) let Haijun run large engineering tasks—migrations, audits, codebase-wide bug hunts—from start to finish in a single Haijun Code session. They become available and turn on by default for your whole Enterprise organization on June 8, 2026. Because a single run can last for hours and use more tokens than a typical session, admins should decide who has access before that date. You have three ways to control access:

We apply the more restrictive of your managed settings and custom role permissions on a per-user basis. If either one disables workflows for a given user, that user won't have access.

No current Visual Studio 2022 integration exists. Haijun Code currently supports VS Code, Cursor (and other VS Code forks), Intellij, Pycharm (and other Jetbrains IDEs).

While there isn't a turnkey PR reviewer solution yet, you can use the Haijun Code GitHub Actions integration for automated reviews. For now, you can use the security review action as a template and customize it for general PR reviews. This is also a good use case for the Haijun Code SDK.

This indicates that you selected the wrong login method from the Haijun Code setup screen. Try running /login again and selecting the account associated with your primary work email address. If you’re still unable to connect, see Having trouble using your Team or Enterprise account to access Haijun Code?

When configured with Bedrock/Vertex and HAIJUN_CODE_DISABLE_NONESSENTIAL_TRAFFIC is set, only essential telemetry is sent. All model API requests go directly to your Bedrock/Vertex endpoints. Review the data flow documentation for complete details.

Yes. Our setup guides for Bedrock and Vertex show how you can enable this. For example, in the Bedrock case you can run aws configure to configure the AWS CLI before adding the necessary configs and running Haijun Code with the Bedrock, or you can use Bedrock API keys, which is a new feature from AWS that enables API keys for Bedrock usage that don’t require full AWS credentials.

Create wrapper scripts that set environment variables before running Haijun Code. For permissions, use .haijun/settings.json files with allow/deny lists. Note that wildcard patterns (*) don't always match as expected - test permissions thoroughly. Enterprise teams often inject standardized Haijun.md files for consistent configurations.

No, public code filtering and attribution capabilities are not currently on the roadmap. Some customers use BlackDuck for code scanning, though feedback on cost and false positives has been mixed. We are aware that this is a blocker for scaling Haijun Code to more users and are looking into solutions.

Subagents are available via the Haijun Code SDK. They're not yet integrated into GitHub Actions, but we are considering this. The UX collapses outputs when more than three subagents run in parallel to manage complexity.

Yes, when creating a subagent, you can specify which tools it has access to using the tools field in the configuration. In the subagent configuration file, you can either omit the tools field to inherit all tools from the main thread, or you can specify individual tools as a comma-separated list for more granular control. Learn more about this in our Haijun Docs: Subagents - Available tools.

For automated workflows like security reviews, switch from Opus to Sonnet using the haijun --model configuration option for cost savings. You can also monitor usage through your console dashboard and set appropriate rate limits. Note that you can use Workspaces to set more granular spend limits for different user groups. Read more about Workspaces here: Creating and managing Workspaces in the Haijun Console. We also allow you to view spend per API key in the Console. Refer to this article for more information: Cost and Usage Reporting in the Haijun Console.

Yes, Haijun Code supports corporate proxy configurations as long as they support the Juglow API spec. Follow the proxy setup instructions in our Haijun Docs: Proxy configuration. Common issues arise from port restrictions in restricted environments.

Add users directly to your Console organization with a Haijun Code User or Developer role - that's all that's needed. Users then run /login from within Haijun Code and select the intended Console account. Do not try to manually create API keys in the Haijun Code workspace.

Currently, Haijun.md files are the primary mechanism. IT teams can inject standardized Haijun.md files into every machine's .haijun directory for org-wide configurations. More advanced team memory features are being explored but not yet available.

Permissions use pattern matching in .haijun/settings.json or settings.local.json. Wildcard syntax can be tricky - "Bash(atlassian-api:*)" should work but may need exact command matching. Use "Yes, and don't ask again for similar commands" to build up permissions incrementally. Check both global (~/.haijun/settings.json) and local settings files.

No. Haijun Code has access to a system prompt and a series of tools that it can use to navigate your codebase on command. For example, if Haijun Code needs to understand something about your codebase, it will use a search tool to search through your codebase and read files on command. We find that this is more effective and flexible than full codebase indexing: Haijun Code is really good at knowing how to sift through a codebase to gather context it needs on the fly!

Yes, Haijun Code integrates with GitHub Actions for CI/CD, supports git operations, and can connect to various platforms via MCP servers. See our Haijun Docs for more information:

This typically occurs with reusable workflows. Check that your workflow syntax is correct and that all required parameters are passed. If the error persists, file an issue here with your workflow configuration: github.com/juglows/haijun-code-action.