[Feature] Support multiple accounts per git provider (e.g. two GitHub accounts)
Nobody has claimed this yet.
Assessment
- Difficulty
- 5/5
- Estimated time
- Over a week
- Newbie friendliness
- 35/100
- Issue type
- Feature
- Clarity
- Mostly clear
- Activity status
- Quiet
- Tech stack
- python
- Domain
- api, authentication, backend
Research direction
Start with openhands/app_server/integrations/provider.py and openhands/app_server/settings/llm_profiles.py to compare the existing token and profile models. Then inspect the settings/secrets routers, git search endpoints, and Settings UI described in the issue. Done means an agreed account model supports multiple provider accounts, preserves legacy single-account storage, and covers account selection across the affected API and UI.
Written by the indexing model from the issue text.
Description
Problem
ProviderHandler and the settings model store exactly one credential per ProviderType:
PROVIDER_TOKEN_TYPE = Mapping[ProviderType, ProviderToken]
(openhands/app_server/integrations/provider.py)
A user with two GitHub accounts (e.g. personal + work, or two orgs that don't share SSO) can only ever have one connected at a time. Connecting a second account overwrites the first — there's no way to keep both and pick per-repo/per-conversation which one to use. Same limitation applies to GitLab, Bitbucket, etc.
I searched existing issues (multiple github accounts, multiple git provider tokens, switch github account, one token per provider, the integrations label, and related closed proposals like #10850 and #15492) and didn't find this specific gap tracked, so filing it separately. Apologies in advance if I missed a duplicate.
Proposed design
The codebase already solved this exact shape of problem for LLM configs — LLMProfiles (openhands/app_server/settings/llm_profiles.py) stores a named dict[str, LLM] plus an active: str | None, with validators that reconcile active against the dict and skip individually-invalid entries on load. Git provider tokens could reuse the same pattern instead of inventing a new one:
class ProviderTokenSet(BaseModel):
"""Named collection of ProviderToken for one provider, plus which one is active.
Mirrors LLMProfiles. Backward compatible with today's bare-ProviderToken
storage via a `mode="before"` adapter.
"""
model_config = ConfigDict(validate_assignment=True)
accounts: dict[str, ProviderToken] = Field(default_factory=dict)
active: str | None = None
@model_validator(mode="before")
@classmethod
def _lift_legacy_single_token(cls, value: Any) -> Any:
# Old on-disk shape for this provider was a bare ProviderToken
# payload (token/user_id/host at the top level). Lift it into the
# new shape transparently so existing users don't lose their
# connection on upgrade.
if isinstance(value, dict) and "accounts" not in value and (
"token" in value or "user_id" in value or "host" in value
):
return {"accounts": {"default": value}, "active": "default"}
return value
@model_validator(mode="after")
def _reconcile_active(self) -> "ProviderTokenSet":
if self.active is not None and self.active not in self.accounts:
object.__setattr__(
self, "active", next(iter(self.accounts), None)
)
return self
def active_token(self) -> ProviderToken | None:
return self.accounts.get(self.active) if self.active else None
PROVIDER_TOKEN_TYPE = Mapping[ProviderType, ProviderTokenSet]
ProviderHandler.get_service becomes the one required call-site change — everything else in that class (get_user, get_repositories, search_repositories, verify_repo_provider, get_authenticated_git_url, etc. — about a dozen methods) still does for provider in self.provider_tokens / self.get_service(provider) and keeps working unmodified, because it's still iterating provider types, not accounts:
def get_service(
self, provider: ProviderType, account: str | None = None
) -> GitService:
token_set = self.provider_tokens[provider]
token = token_set.accounts[account] if account else token_set.active_token()
if token is None:
raise AuthenticationError(f"No active account configured for {provider}")
service_class = self.service_class_map[provider]
return service_class(
user_id=token.user_id,
external_auth_id=self.external_auth_id,
external_auth_token=self.external_auth_token,
token=token.token,
external_token_manager=self.external_token_manager,
base_domain=token.host,
)
The optional account param is additive — every existing caller that doesn't pass it keeps resolving to today's single/active token.
API surface (sketch, not implemented)
Following the shape /api/v1/settings/profiles already established for LLMProfiles:
| Method | Path | Purpose |
|---|---|---|
| GET | /api/v1/secrets/git-providers/{provider}/accounts |
list connected accounts for a provider |
| POST | /api/v1/secrets/git-providers/{provider}/accounts/{name} |
add/update a named account |
| DELETE | /api/v1/secrets/git-providers/{provider}/accounts/{name} |
remove a named account |
| POST | /api/v1/secrets/git-providers/{provider}/accounts/{name}/activate |
switch which account is used by default |
The existing POST /api/v1/secrets/git-providers (bare provider→token) would keep working by writing to accounts["default"] + active="default", so nothing breaks for single-account users.
What this issue is not
I deliberately stopped at the model + one method, not a full PR. This type touches the settings/secrets routers, every git search endpoint (git/repositories/search, git/branches/search, users/git-info), and presumably the Settings UI for account selection — real work, but I wanted to check the shape of the approach first, especially since #10850 and #15492 suggest the settings/integrations area is under active redesign. Happy to build this out (backend + a minimal UI) if the direction looks right, or to adjust if there's a preferred design already in flight.
Open questions for maintainers
- Is per-provider "named accounts with one active" (mirroring
LLMProfiles) the right mental model, or is per-repository/per-conversation account selection (no global "active" account) closer to what's planned for the settings/integrations redesign? - Should account names be free-text (like LLM profile names) or should the connect flow auto-derive a label from the authenticated username (avoids "personal"/"work" bookkeeping the user has to invent)?
- Dominant language
- Python
- Stars
- 9
- Forks
- 16
- Avg merge
- 7d 4h
- Merged PRs (30d)
- 2
Getting set up
- Ships a Dockerfile or Docker Compose file
- No pull request template
- No contributing guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
More from OpenHands/sandbox-server
-
Difficulty 5/5 Over a week Newbie friendliness 25/100
OpenHands/sandbox-server#23 ·
-
Difficulty 5/5 Over a week Newbie friendliness 35/100
OpenHands/sandbox-server#18 ·
-
Difficulty 1/5 Under an hour Newbie friendliness 20/100
OpenHands/sandbox-server#12 ·
-
sandbox-server Gateway and V1 routing for self-hosted deployments + agent-canvas backend supportOpen
Difficulty 5/5 Over a week Newbie friendliness 30/100
All issues in OpenHands/sandbox-server
Similar issues
-
[Bug]: reme CLI crashes when the running server used a relative config pathPossibly taken @nurlanmnn claimed this today. Open
Difficulty 2/5 1-3 hours Newbie friendliness 88/100
agentscope-ai/ReMe#770 ·
Maintainers usually reply within 1 day
-
Pi: paused model calls are reported as "not signed in" when the model's provider comes from an extensionPossibly taken @balcsida claimed this today. Open
Difficulty 2/5 1-3 hours Newbie friendliness 73/100
microsoft/TokenEconomizer#288 ·
Maintainers usually reply within 1 day
-
Difficulty 1/5 Under an hour Newbie friendliness 67/100
sickn33/agentic-awesome-skills#2437 ·
Maintainers usually reply within 1 day
-
site-bug triage
Difficulty 2/5 1-3 hours Newbie friendliness 72/100
Maintainers usually reply within 1 day
-
Difficulty 2/5 1-3 hours Newbie friendliness 78/100
ethereum/consensus-specs#5846 ·
Maintainers usually reply within 1 day