One account, every member
Manage everyone under one enterprise account; allocate quotas and cap usage by role.
Your company shouldn’t be locked to one AI
One OpenAI-compatible gateway for DeepSeek, GPT, Claude, Gemini and Grok. One account, USD billing and enterprise controls — with DeepSeek V4.1 Flash ready first.
https://api.opencaply.com/v1
OpenAI-compatiblePay as you go · No monthly fee · Switch models anytime
One integrationFree switchingAccounts & permissionsQuota transfersCost visibilityAudit trail
One enterprise account · one AI control layer · every model and tool behind it
Anthropic
Grok
QwenAI VENDOR INDEPENDENCE
Choosing the wrong AI shouldn’t mean starting over
Models change, prices change, capabilities change. Betting your whole AI stack on a single vendor is the real risk.
Bet on one vendor
Every switch is a rebuild
- Pick a vendor and sign
- Roll it out company-wide; workflows and data follow
- Six months later, another model fits better
- Migration costs too much to move
- Locked in
Run on OpenCaply
Switching models is not switching systems
- One account for every leading model
- One set of keys, permissions and quotas
- People keep the tools they already use
- Switch whenever you want, revert anytime
- Account, members, permissions, records and assets all stay
Free to switch · Always reversible · Assets and memory preserved
Smart routing
Channel-level smart failover
When an upstream fails, the gateway switches channels in milliseconds — invisible to your business.
High availability
Redundant upstream channels — single points of failure don’t stop service.
Low cost
Routing by weight and price policy keeps spend transparent and under control.
Low latency
Nearest and fastest-responding channels get priority.
Guaranteed
Failover is fully automatic — no manual intervention.
CONTROL
Let your people use AI — keep control in the company
Personal ChatGPT accounts, ad-hoc developer keys, finance with no idea of the spend, IT with no idea who is using what. That shadow AI is the real risk.
Keys and permissions
Every member gets their own key — revocable anytime, with IP allowlists and allowed-time windows.
Spend and quotas
See consumption by member, key and model; usage stops automatically at the limit, with cost visible live.
Audit and traceability
Who used which model, when, and at what cost — request-level detail, fully searchable.
Shadow AI won’t disappear on its own. But it can be seen.
WORK SUMMARY · EMPLOYEE SUMMARY
Not just what you spent — what got done.
Based on your company’s own AI memory, work summaries are organized by employee and date. This is not screen monitoring; it turns AI collaboration that already happened into reviewable, handover-ready work records.
Review by person and day
Choose a member and date range; see what moved forward, where it stalled, and what comes next.
From conversations to conclusions
AI reads your company’s encrypted memory and condenses scattered conversations into a readable summary, not raw chat logs.
Boundaries before features
No screen capture, no keystroke logging; the key stays with the owner and the server discards it after use; analysis is billed to the company and generated on demand.
Employee work summary
Mon 10/13Client interviews
Customer research · competitive analysis
Completed 3 customer interviews and identified two recurring objections.
Compared competitor pricing pages and produced 5 revision suggestions.
Tue 10/14Solution review
Design review · prototype
Reviewed the new onboarding flow with design and agreed on 2 key changes.
Wed 10/15Data review
Weekly metrics review
Summarized last week’s conversion data and located the drop-off point on the signup page.
Reads your encrypted AI memory only · Private key stays on your device
No screen monitoringNo keystroke loggingOnly your company’s AI memory
RISK & SECURITY
The company controls the AI — not the other way round
From data flow to content boundaries, from key management to upstream reliability, every claim below maps to a mechanism you can verify.
Data and content control
What leaves the company is your call
Controlled data flow
Request content is forwarded and metered only — never used for model training, never sold to third parties.
Content security
Create guard profiles (security groups) bound to API keys to block classified data, malicious code and phishing in real time — with full detection logs.
Access defenses
TLS across all traffic; per-key IP allowlists and allowed-time windows; email-code login and team security policies guard account access.
Reliability and credential safety
Upstream failures don’t stop your business
Keys stored as hashes
An API key is shown once and stored only as an irreversible hash — it cannot be recovered from the hash.
Credentials encrypted
Channel credentials are sealed with AES-GCM — never stored or logged in plaintext, safe to share across the team.
Key pools and auto-failover
Multiple upstream key pools per model; a failing key or channel is switched automatically and switched back on recovery.
MEASURE
Your spend is visible. The value should be too.
From AI spend to business output, OpenCaply puts both on the same line so that “is AI worth it?” has an answer.
Spend
Break every AI dollar down by member, key, model and channel — live and drillable.
Activity
Attribute each call to real work: client analysis, quotes, technical proposals, support replies.
Outcome
Convert completed work into hours and payroll saved, so you can judge whether the spend pays off.
Example · one member · this monthIn progress
AI cost
$283
Work completed
- 37 client analyses
- 12 quotes
- 28 technical proposals
Est. hours saved
74 hours
Spend and usage detail is live today; output attribution and ROI reports are in progress.
KEEP Planned
People leave. Their capability stays.
Models change, tools change — but the judgment and experience your company accumulates shouldn’t drain away with them. OpenCaply has AI continuously learn how your people work and preserves it as inheritable org memory — people leave, judgment stays; new joiners take over on day one.
01
Capture
Every working moment, captured automatically
AI keeps recording how your people decide and solve problems, building a methodology library unique to your team. People leave — their judgment stays.
02
Clone
An AI twin for every key role
Distill the capability profiles of core employees. Successors simply ask the AI twin — answers come from years of real-world experience.
03
Inherit
Zero gaps in critical handovers
Successors are up to speed on day one — the AI always knows "how this used to be done". No more relying on word of mouth.
Memory stays in the teamNew hire takes over
Org memory is encrypted — 100% secure
All memory data is stored encrypted, and the decryption key is held solely by the enterprise owner — neither the platform nor any third party can read it.
Built on a routing layer you can run in production
Baseline figures for our upstream channels — actual values vary by upstream and time, and every request is routed live.
Multiple upstream channels compete on price under continuous quality monitoring, so every request gets the best available rate — live pricing for every model is published on this page.
Slow routes are pulled from rotation automatically, and the nearest, fastest nodes get priority — time to first token is monitored continuously.
Multi-key pooling and channel-level failover back a production-grade SLA: automatic failover across providers, 24/7, no downtime.
Model marketplace
One API, the entire model marketplace
Official models and enterprise sellers side by side, with input and output priced separately. Sign up and call everything with one API key.
DeepSeek V4.1 FlashDeepseekOfficialInput / 1K tokens
$0.00039
Cached / 1K
$0.0000078
Output / 1K tokens
$0.00156
Qwen 3.8 MaxQwenOfficialInput / 1K tokens
$0.002
Cached / 1K
$0.00025
Output / 1K tokens
$0.006
Grok 4.7GrokOfficialInput / 1K tokens
$0.002
Cached / 1K
$0.0005
Output / 1K tokens
$0.006
Gemini 4 ArgonGoogleOfficialInput / 1K tokens
$0.002
Cached / 1K
$0.0002
Output / 1K tokens
$0.01
Claude Opus 5.5AnthropicOfficialInput / 1K tokens
$0.004
Cached / 1K
$0.0002
Output / 1K tokens
$0.02
GPT 6 AstraOpenAiOfficialInput / 1K tokens
$0.01
Cached / 1K
$0.001
Output / 1K tokens
$0.05
GPT 6.1 SolOpenAiOfficialInput / 1K tokens
$0.002
Cached / 1K
$0.0001
Output / 1K tokens
$0.01
GLM 5.3GLMOfficialInput / 1K tokens
$0.0014
Cached / 1K
$0.00026
Output / 1K tokens
$0.0044
Claude Fable 5.1AnthropicOfficialInput / 1K tokens
$0.01
Cached / 1K
$0.00025
Output / 1K tokens
$0.05
Enterprise offers are reviewed before listing; availability is confirmed in the console.
GENERATION PRICING
Image and video, billed by usage
Images are billed per image and videos per second, using the same gateway, API key, balance and permissions as chat models.
Image generation
Gemini 3.1 Flash ImageGooglePer image
512x512$0.045/ image
1024x1024$0.067/ image
2048x2048$0.134/ image
4096x4096$0.151/ image
Gemini 3 Pro ImageGooglePer image
1024x1024$0.134/ image
2048x2048$0.134/ image
4096x4096$0.24/ image
Nano Banana 2.1GooglePer image
1024x1024$0.0336/ image
Video generation
Veo 3.1GooglePer second
Standard$0.4/ second
Veo 3.1 FastGooglePer second
Standard$0.15/ second
Veo 3.1 LiteGooglePer second
Standard$0.05/ second
Final price depends on the model and size used; video jobs settle by actual duration.
Every token, accounted for
Pay as you go — no monthly fee, no minimum top-up. Balance never expires, and every request is itemized in real time.
Balance never expires · Top up anytime
Quick start
100% protocol compatible — change one Base URL line and ship to production;
or download the OpenCaply desktop app and connect Codex / Claude Code in one click.
1from openai import OpenAI2 3client = OpenAI(4 base_url="https://api.opencaply.com/v1", # the only line that changes5 api_key="<OPENCAPLY_API_KEY>",6)7response = client.chat.completions.create(8 model="DeepSeek V4.1 Flash",9 messages=[{"role": "user", "content": "Hello"}],10)11print(response.choices[0].message.content)
Point the base URL at OpenCaply, drop in your API key — the rest of your code stays the same.
Get started, pay as you go
An API key on sign-up — change one line of base_url and start calling.
Pay as you go · No monthly fee · Balance never expires
FAQ
FAQ
OpenCaply is an enterprise AI control layer: every leading model behind one account, one place to govern, and the freedom to switch — account, members, permissions, keys and history all stay when you change models.
Fully compatible. The interface matches OpenAI’s format, and official SDKs work without modification.
By token usage, with input and output priced separately. No monthly fee and no minimum spend. Balance never expires; top up anytime.
API keys are stored as hashes only, and upstream credentials are encrypted. Request content is forwarded and metered only — never used for training.
Sign up free for an API key and point your client’s base_url at the gateway — done in three steps.
Team spaces are supported: member management, quota transfers and role permissions under one account system, with unified billing.
No. Switching is a console change — account, members, permissions, keys and usage history all stay, and your code doesn’t change at all.
No. It does not capture screens or log keystrokes; it only reads your company’s encrypted AI memory. The analysis key stays with the owner and is discarded by the server after use. Summaries are generated on demand for review and handover.