Screenshots 1
1endpoint
Summary
Every time you swap providers — OpenAI to Anthropic, Anthropic to DeepSeek — you're touching client code, updating auth, rewriting endpoints. 1endpoint puts a single base URL in front of all of them so the swap is a model ID change, not a refactor.
The gateway accepts OpenAI and Anthropic request shapes natively, so existing client code points at one base URL and keeps working. Prompt caching is baked into the billing model: by message 12 of a conversation, the vendor's published illustration shows costs running at roughly 3.5× less than uncached, because only the new tokens pay full price inside a rolling 30-minute session window. Token-level spend tracking records usage per successful response with no blended platform fee hidden in the rate. The ceiling appears when a session lapses — the window resets, the first request pays full price again, and teams with long-gap workflows lose most of the cache benefit. There is no self-hosted option, so every request leaves your infrastructure.
Bottom line: Pick this for high-message-count chat workloads where cache hit rates are predictable and latency to an external gateway is acceptable — abandon it when your compliance policy requires data to stay on-premises or your sessions are too sporadic to hold a cache window.
Pricing Plans
Usage-BasedUsage-based
Pay per token; 1,000 credits = $1; input starts at $0.0300 per 1M tokens
- Per-model input/cached/output rates
- Automatic cache discounts
- No platform fees
View full pricing on 1endpoint.dev →
Pricing may have changed since last verified. Check the official site for current plans.
Community Performance Report Card
No community ratings yet. Be the first to rate this tool!
Pros
Sign in to edit- OpenAI and Anthropic request format compatibility out of the box, so existing client libraries need no changes — a provider switch is one model ID edit, not a code change.
- Token-level cost breakdown separates input, cached input, and output with no blended platform surcharge, so cost attribution per conversation is exact rather than estimated.
- Committed cache rate bands on marked models mean long conversations accrue a predictable discount rather than a best-effort one — the vendor publishes the band width and the conditions, so you can model costs before committing.
- A single set of credentials and one base URL covers the full model catalog, so teams with multiple projects using different models avoid managing separate API keys and authentication flows per provider.
Cons
Sign in to edit- The 30-minute cache window resets on any gap between requests — sporadic or async workloads (batch jobs, low-traffic overnight sessions) pay full input price on the resuming request, erasing the cost advantage that justifies routing through the gateway at all.
- No self-hosted deployment exists; all traffic passes through the vendor's infrastructure. Teams under HIPAA, SOC 2 data-handling restrictions, or internal policies that prohibit third-party API proxies hit a hard wall with no workaround — this is the condition that sends those teams to direct-provider endpoints or a self-hosted inference proxy instead.
- Unmarked models in the catalog carry no committed cache rate, meaning cache behavior on those models is present but unguaranteed — teams building cost projections around cache efficiency on unmarked models are working from incomplete data and will need to validate empirically.
About
- API Available
- Yes
- Self-Hosted
- No
- Last Updated
- 2026-09-08T20:07:59.965Z
Best For
Who it's for
- Developers needing OpenAI/Anthropic compatible endpoints
- Cost-sensitive inference at scale
- Workloads benefiting from automatic cache discounts
What it does well
- Unified access to multiple models without client changes
- High-volume workloads with token-level cost tracking
- Applications requiring prompt caching for long conversations
Add notes, reviews, and benchmarks so the next visitor gets a clearer picture.
Compare 1endpoint
Spotted incorrect or missing data? Join our community of contributors.
Sign Up to ContributeFrequently Asked Questions
- Is 1endpoint free?
- 1endpoint is a paid tool. No permanent free tier is offered.
- Is 1endpoint open source?
- No — 1endpoint is a closed-source tool. Source code is not publicly available.
- Does 1endpoint have an API?
- Yes. 1endpoint exposes a developer API. See the official documentation at https://1endpoint.dev for details.
Curated lists that include this category
Switching LLM providers mid-project normally means touching authentication, endpoint URLs, and sometimes request shapes. 1endpoint is a paid API gateway that presents one base URL accepting OpenAI chat completions, OpenAI responses, and Anthropic messages formats — the model ID in the request is the only thing that changes when the workload changes. The catalog spans flash-class and reasoning-class models from GLM, GPT, DeepSeek, Gemini, Minimax, Kimi, and Opus families, priced per million tokens with input, cached input, and output billed as separate line items.
The differentiating mechanism is prompt caching with a committed rate band on marked models. Because each message in a conversation resends all prior context, the gateway charges full price only for tokens not already cached. The vendor’s published illustration — 12 messages, each adding 10,000 input and 1,000 output tokens — puts the cached cost at roughly $1.06 per 1,000 requests versus $3.70 without cache at GLM 5.3 Flash rates. The cache window runs 30 minutes, restarted by each successful request; a gap beyond that resets the session and the first new request pays full price.
The tool fits teams running high-volume chat or RAG pipelines where sessions are dense and continuous — customer support agents, document chat interfaces, coding assistants with long context. It breaks for workloads where sessions are sparse or asynchronous: the cache window closes, the economics flatten toward standard per-token pricing, and the routing abstraction stops justifying the external hop. There is no self-hosted deployment path, so teams under data-residency or compliance constraints that prohibit external API proxies have no workaround.
Integration is a base URL change to https://1endpoint.dev/api/v1 and an API key swap — POST /chat/completions, POST /responses, and POST /messages all work against the same host. Spend tracking records usage after each successful response at the currently configured rates, with no retroactive adjustment; the blended reference rate shown in the catalog assumes 70% cache hit and output at 25% of input volume, which is a comparison reference, not a billed construct.