Notify authors
Is your feature request related to a problem? Please describe.
ClawCodex is an open-source, Python-native rebuild of Claude Code that frees agentic coding from a single model vendor: it wires providers such as Anthropic, OpenAI, Z.ai, MiniMax, OpenRouter and DeepSeek into one ~/.clawcodex/config.json (default_provider plus per-provider base_url, api_key, default_model). For developers who want a Claude Code-grade runtime they can point at any model stack, that portability is the core value.
The payoff is clearest in long sessions: ClawCodex keeps request prefixes byte-stable so DeepSeek's prompt cache bills cache-hit input at about $0.022 per 1M tokens — "run long agentic coding sessions for pennies" — and the /eco toggle measurably cuts Bash-output tokens by roughly 80%. Sustained, cost-sensitive agentic work benefits from options that preserve that efficiency and add reliability, not just more endpoints.
Describe the solution you'd like
I'd like to propose adding OrcaRouter as an optional provider for ClawCodex. It would not replace or alter any existing provider block — just one more entry a user can opt into, like openrouter or deepseek.
OrcaRouter exposes an OpenAI-compatible API and uses standard API-key authentication, so the natural integration point is the provider abstraction ClawCodex already has: a config.json provider block whose base_url points at OrcaRouter's endpoint, with api_key and default_model set alongside, and model names carried over the OpenAI-compatible wire protocol. Nothing here is implemented or tested yet — this issue is to gauge maintainer interest before any PR.
For ClawCodex users, the most relevant capabilities are:
- Automatic model routing and provider failover — a long agentic run can be routed across upstream providers and fall back when one is slow, rate-limited, or down.
- Prompt caching across turns, which fits ClawCodex's byte-stable prefixes and its token-efficiency focus.
- Usage tracking, budgets, and team access controls — useful when a team shares an install or wants to see what long sessions cost.
I'm an engineer on the OrcaRouter team. OrcaRouter is already used in the open-source ecosystem — goose, OpenCode, models.dev, and promptfoo, among others — with more projects at https://www.orcarouter.ai/built-with.
One transparent disclosure: OrcaRouter runs an optional open-source partner program where approved OSS projects can receive a 5% revenue share from OrcaRouter usage attributed to their integration. Participation is not a prerequisite, and I'd be glad to follow any disclosure or governance rules ClawCodex wants to apply.
Describe alternatives you've considered
Users who want many models can already add OpenRouter or another OpenAI-compatible provider directly. OrcaRouter's difference is the routing, failover, and caching layer behind a single key; if maintainers prefer to keep the provider list unchanged, closing this issue is reasonable.
Additional context
Happy to provide an OrcaRouter API key for testing and, if you're open to it, to submit an implementation PR per CONTRIBUTING.md. Thanks for building a provider-neutral agent runtime.
Notify authors
Is your feature request related to a problem? Please describe.
ClawCodex is an open-source, Python-native rebuild of Claude Code that frees agentic coding from a single model vendor: it wires providers such as Anthropic, OpenAI, Z.ai, MiniMax, OpenRouter and DeepSeek into one
~/.clawcodex/config.json(default_providerplus per-providerbase_url,api_key,default_model). For developers who want a Claude Code-grade runtime they can point at any model stack, that portability is the core value.The payoff is clearest in long sessions: ClawCodex keeps request prefixes byte-stable so DeepSeek's prompt cache bills cache-hit input at about
$0.022per 1M tokens — "run long agentic coding sessions for pennies" — and the/ecotoggle measurably cuts Bash-output tokens by roughly 80%. Sustained, cost-sensitive agentic work benefits from options that preserve that efficiency and add reliability, not just more endpoints.Describe the solution you'd like
I'd like to propose adding OrcaRouter as an optional provider for ClawCodex. It would not replace or alter any existing provider block — just one more entry a user can opt into, like
openrouterordeepseek.OrcaRouter exposes an OpenAI-compatible API and uses standard API-key authentication, so the natural integration point is the provider abstraction ClawCodex already has: a
config.jsonprovider block whosebase_urlpoints at OrcaRouter's endpoint, withapi_keyanddefault_modelset alongside, and model names carried over the OpenAI-compatible wire protocol. Nothing here is implemented or tested yet — this issue is to gauge maintainer interest before any PR.For ClawCodex users, the most relevant capabilities are:
I'm an engineer on the OrcaRouter team. OrcaRouter is already used in the open-source ecosystem — goose, OpenCode, models.dev, and promptfoo, among others — with more projects at https://www.orcarouter.ai/built-with.
One transparent disclosure: OrcaRouter runs an optional open-source partner program where approved OSS projects can receive a 5% revenue share from OrcaRouter usage attributed to their integration. Participation is not a prerequisite, and I'd be glad to follow any disclosure or governance rules ClawCodex wants to apply.
Describe alternatives you've considered
Users who want many models can already add OpenRouter or another OpenAI-compatible provider directly. OrcaRouter's difference is the routing, failover, and caching layer behind a single key; if maintainers prefer to keep the provider list unchanged, closing this issue is reasonable.
Additional context
Happy to provide an OrcaRouter API key for testing and, if you're open to it, to submit an implementation PR per CONTRIBUTING.md. Thanks for building a provider-neutral agent runtime.