Use Claude Code with HIPS
hi claude starts Claude Code on HIPS models, signed in as you, with no API key to copy: hi sets HIPS up for that session only and leaves Claude Code's own settings as they are.
curl -fsSL https://hips.com/hi.sh | shhifindsclaudeon your computer, checks that its release is one HIPS has tested on your system, and names its vendor's installer when it is missing.- The first time, it signs you in: it shows a code and a link, and you approve it in the Console.
- It takes a model fit for coding from your workspace's catalog, or the one you name with
--model, and gives Claude Code HIPS's address and your sign-in through Claude Code's own credential setting, speaking Anthropic's Messages API. - It turns off Claude Code's own telemetry, and whatever would send your session to its vendor, for that session.
Code with HIPS is the overview; Coding agents has every option of hi claude.
Starts in Accept edits (acceptEdits): it edits files and runs common file commands without asking, and asks before anything else. Its auto mode, in which a model reviews each action, is not supported by hi yet: hi leaves it out of Shift+Tab, and a session you start in auto mode, with --permission-mode auto or auto as permissions.defaultMode in your settings, starts in Accept edits instead. Any other mode you name is yours.
Auto mode, where a model reviews or approves an agent's actions for you, is not supported by hi yet; Coding agents has every agent's starting mode.
| System | Tested releases |
|---|---|
| macOS | 2.1.314 to 2.1.325 |
| Linux | 2.1.319 to 2.1.325 |
| Windows | 2.1.322 to 2.1.325 |
- System
- macOS
- Tested releases
- 2.1.314 to 2.1.325
- System
- Linux
- Tested releases
- 2.1.319 to 2.1.325
- System
- Windows
- Tested releases
- 2.1.322 to 2.1.325
From the signed list of tested versions of 2026-11-14, 01:09 UTC (sequence 7), joined with the releases built into hi 0.1.3. hi reads the signed list each day, so it may start a newer tested release than this page shows; any other release is refused with the tested ones named.
In the per-model agent tests' record of 2026-11-14, 05:25 UTC (sequence 10): Passed in the Community pool with gpt-oss-20b, granite-4.2-8b and qwen2.5-7b-instruct, on Claude Code 2.1.325. The gateway refused the pairs of two other models, and the rest found no free capacity when they ran; both run again with the next tests. The Console's Models page shows each model's result as the record holds it.
For people who do not use hi. Named by the source specification as a candidate where compatibility can be established. Started on HIPS by hi claude; its manual setup is unvalidated.
| Status | Tool and version tested | Validated on |
|---|---|---|
| unvalidated | none | none |
- Status
- unvalidated
- Tool and version tested
- none
- Validated on
- none
Not tested. When the row is validated, this section states where the tool reads its base URL, key and model (environment, file, flag or UI), and which of those the setup helper uses for a session and for a persistent configuration.
Not tested. When validated: how the tool sends the key, that HIPS accepts it as Authorization: Bearer, and that the helper never prints a secret. The key is read as HIPS_API_KEY and kept on the machine's own secret store or environment.
Not tested. When validated: the backup the helper takes before a persistent change, and the selective undo that restores the previous provider, model mapping and defaults without touching unrelated providers.
- Tool and SDK versions: none tested.
- Configuration surface: not tested.
- Authentication: not tested.
- Model mapping: not tested.
- Streaming: not tested.
- Complete tool-call round trips: not tested.
Note
The status is unvalidated Not decided yet. A page moves past unvalidated only with a dated validation record; the record's date and test results replace the "not tested" lines above.
Related sheets
- Sheet
- 10 of 20
- Detail
- B
- Updated
- 2026-11-14