Reserve budget before the call
Conservative reservations check available budget before provider dispatch. Customer, project and scoped-key limits keep concurrent work within configured spending rules.
Oryndra · Pre-launch
Open for buyer evaluation · Pilot planned
Oryndra
LLM Spend Controller
Customer-hosted AI access with spending rules
Control AI spending, onboard your team, trace usage to individual requests, and recover accounts with safeguards for saved conversations—all in one customer-hosted gateway and chat workspace.
Before a request reaches a model
A team member or connected app makes a request.
Access, budget, and approval rules are checked first.
An eligible request is sent through the customer-hosted gateway.
The response and configured usage record are saved.
Bring the gateway, employee workspace and administrative controls together—from the first invitation to the audit record behind a request.
Reserve budget before dispatch and trace costs from each person to each request.
Onboard with shared presets and recover accounts with safeguards for earlier chats.
Reopen saved conversations, export supported media and use checkpointed workflows.
01 / Spending & accountability
Set the rules before dispatch, then follow usage from the employee or application down to the individual request.
Conservative reservations check available budget before provider dispatch. Customer, project and scoped-key limits keep concurrent work within configured spending rules.
Require human approval above a per-request threshold. Approval authorizes the request while customer, project and key budgets still apply.
Control allowed models, expiry, request rate and concurrency. Separate Users and Client keys pages offer searchable directories, key details, revocation and reusable key presets.
Filter by employee, API user, customer, project, provider, model, status or date. Open a provider card, drill into a request and export the matching usage as CSV.
Inspect the saved rate snapshot, input/output tokens and reported cache or reasoning subsets. Accounted costs and unresolved budget holds stay separate; later pricing edits do not rewrite earlier request rates.
Search audit history by person, event and date. A server-synchronized UTC clock sits beside reporting; daily exports use UTC dates and search filters label their browser-time offset.
02 / Team access & recovery
Give employees a consistent workspace policy, with individual sign-in, storage allowances and explicit recovery safeguards.
Save shared model, budget and storage policies. Invite up to 50 employees per batch and download their unique private links as CSV. Readable assigned usernames are verified and read-only during setup.
Choose link validity from 60 minutes to 48 hours, also bounded by generation expiry. One-time token links verify the assigned account; enrollment returns to normal password and Authenticator sign-in.
Set a 16–1024 MiB allowance per employee, with 64 MiB as the default. Saved text and supported media share the quota, making storage policy an explicit onboarding choice.
Reset a password while retaining the existing Authenticator, replace an Authenticator while retaining the password, or recover both. Each recovery revokes existing sessions and retains the employee identity.
Full recovery locks earlier chats until an authorized server operator unlocks them. Staff browser roles cannot read or unlock employee archives. Full recovery, operator unlocks and post-recovery access leave audit records.
Store provider keys with AES-256-GCM encryption and replace credentials without reading back existing secrets. Vault wrapping-key rotation is a separate operation. Authorized host operators remain trusted.
03 / Workspaces & integrations
Saved conversations, portable exports and checkpointed workflows support work beyond a single browser tab.
Encrypted conversation history persists across refresh and sign-in. Administrators have their own saved-chat list and reconnect a scoped key before continuing. Generation expiry preserves saved employee history.
Use documented image and audio inputs with compatible routes. User-reviewed conversation summaries help manage longer conversations while keeping the user involved in what is carried forward.
Export as TXT, JSON or self-contained HTML for offline reading. JSON and HTML can include supported media, keeping the conversation useful outside the workspace.
Run ordered model steps with budgets and checkpoints. Opt-in recovery can resume eligible steps after confirmed provider quota rejection; uncertain or interrupted outcomes can require review.
Configure routes and opt-in fallback paths, subject to provider support. Connect applications through the gateway API or cooperative Python SDK, with customer and project accounting boundaries.
Google Workspace SSO supports explicitly authorized administrator identities. Configured Slack or Telegram endpoints can receive budget and security notices through the delivery worker, with retry handling.
Commercial acquisition & licensing
Built as a head start for AI agencies, managed IT providers (MSPs), and B2B software teams looking to rebrand, integrate, and build on a working product foundation.
Option A · Full commercial control
Acquire the Oryndra product foundation for buyer rebranding and commercialization, with final transfer terms documented in the signed agreement.
Option B · Custom brand deployment
An alternative path for AI consultancies, managed service providers, and enterprise software teams interested in deploying Oryndra under their own brand. White-label scope and commercial terms are discussed separately.
The acquisition and white-label paths are alternatives, not simultaneous rights grants. Third-party components remain subject to their own licences; definitive scope and transfer terms require a signed agreement.
Recorded pre-launch evidence
The product walkthrough uses actual product UI captures with synthetic demo data. It is a narrated presentation, not a live interactive recording.
A recorded Groq request matched 80 input and 43 output tokens to the provider dashboard, with $0.295 settled at configured test prices. This was a free-tier call with simulated nonzero pricing.
A recorded paid Gemini call preserved 31 input and 956 output tokens from the incoming API response, including reported thinking. The configured-rate calculation matched the settled $0.003609.
This verifies captured API counters and configured arithmetic. Provider dashboard/log differences are documented separately; provider invoice reconciliation is not established.
Owner-run restart checks preserved the recorded ledger. Employee tests covered password-only and MFA-only recovery with saved chats retained, plus full-reset archive locking, audited operator unlock and an unaffected second account.
A 13-part buyer walkthrough and a verified source package are available for evaluation, with updated setup guides, 23 documented screenshots and file-integrity records.
These are seller-prepared records, not an independent security certification or third-party audit.
Turnkey asset transfer
Materials are organised to help a buyer review, rebrand, and plan deployment with their product and engineering teams.
Gateway, employee chat portal, and cooperative Python SDK.
Test suite, recorded acceptance evidence, and walkthrough materials.
Step-by-step commands and 23 screenshots covering setup, onboarding, spending and recovery.
Command-based troubleshooting, third-party notices, CycloneDX/SPDX inventories and IP documentation.
Ready for review
Ask questions, review the walkthrough, or discuss either commercial path.