Pricing & usage
Accly paid plans use weighted request units for customer-facing usage. Request units are daily usage capacity for the Gateway API and hosted app model usage. They are not transferable and cannot be redeemed for money.
Request units
Each paid plan includes a daily request-unit allowance:
| Plan | Daily request units | Reset |
|---|---|---|
| Starter | 400 | Midnight UTC |
| Pro | 1000 | Midnight UTC |
| Max | 4000 | Midnight UTC |
One API call does not always equal one unit. Lightweight model requests can consume one unit. Heavier models, larger context windows, long outputs, cached context reads, tool-heavy requests, and agentic coding sessions can consume more units.
Request units do not roll over. Unused daily units expire at the daily reset.
How units are consumed
Accly meters successful model usage and converts it into request units. The exact amount can vary by:
- selected model
- input tokens and output tokens
- context size
- cached context usage
- request shape, such as tools, images, or long outputs
- route availability and model capability support
You can monitor request-unit usage in the Accly app. The usage view shows units used, units remaining, RPM, request outcomes, model activity, and operational errors.
No silent model swaps
Accly does not silently replace the model you selected with an unrelated cheaper model.
If a model route is degraded, delayed, unavailable, or economically unsafe to serve, Accly can limit, queue, or temporarily pause that route instead of changing the model behind your back. Model availability can still change because provider routes and beta access can change.
Free trial allowance
Free is for trying Accly in the hosted app. Free does not support Gateway API keys.
The free trial allowance is one-time, does not reset daily, and is shown in the app as trial state. When the free trial allowance is exhausted, new free model requests are blocked until you upgrade.
The current error code for this state is free_balance_exhausted. The name is legacy and may be renamed later; the user-facing behavior is a free trial allowance being exhausted.
Promotional benefits
Promotional codes can unlock trial access, discounts, invite-only access, temporary plan benefits, or extra usage allowance depending on the campaign.
Promotional benefits do not remove model tier rules, request-unit limits, RPM, request-count guards, subscription state, abuse controls, or campaign-specific limits. See Redeem Codes.
RPM
RPM means requests per minute. It controls short-term request speed.
| Plan | Gateway API access | RPM |
|---|---|---|
| Free | No | Hosted app only |
| Starter | Yes | 35 |
| Pro | Yes | 65 |
| Max | Yes | 65 |
If you exceed RPM, the gateway returns rate_limit_exceeded. Wait for the short reset window before retrying.
Request-count guard
Request units are the main public daily capacity. Accly also keeps a daily request-count guard to prevent unusual bursts of many tiny requests.
| Plan | Daily request-count guard |
|---|---|
| Free | 0 API requests |
| Starter | 400 |
| Pro | 1000 |
| Max | 4000 |
When this guard is exhausted, the gateway returns daily_limit_exceeded.
Units vs requests vs RPM
These controls protect different parts of usage:
| Limit | What it controls | Example |
|---|---|---|
| Request units | Weighted daily model usage capacity | A heavy coding-agent request can consume multiple units |
| Request-count guard | Total API request attempts for the day | Many small requests can exhaust the daily guard |
| RPM | Short-term request speed | A burst can hit RPM even with units remaining |
Model tiers
Plans also control model tiers:
| Plan | Tiers |
|---|---|
| Free | Basic |
| Starter | Basic, Advanced |
| Pro | Basic, Advanced, Thinking, Beta |
| Max | Basic, Advanced, Thinking, Beta |
If you request a model tier your plan cannot use, the gateway returns model_not_allowed.
API keys
API keys are available on paid plans.
| Plan | API keys |
|---|---|
| Free | No |
| Starter | Up to 10 |
| Pro | Up to 10 |
| Max | Up to 10 |
Choose the key type based on your client:
openaifor OpenAI-compatible tools and Codexanthropicfor Anthropic-style clientsgooglefor Google/Gemini-style clientsuniversalfor OpenAI-compatible multi-model routing
When limits are hit
| Situation | What happens |
|---|---|
| Free trial allowance exhausted | New Free requests are blocked with free_balance_exhausted |
| Request units exhausted | New paid requests are blocked with request_units_exceeded |
| Request-count guard exhausted | New requests are blocked with daily_limit_exceeded |
| RPM exceeded | Request is blocked with rate_limit_exceeded |
| Too many concurrent requests | Paid requests wait briefly, then can fail with concurrency_queue_timeout |
| Model tier not allowed | Request is blocked with model_not_allowed |
During the request-unit migration, some older clients or cached errors may still refer to budget_exceeded. Treat that as a legacy usage-limit error and check the Accly app for current request-unit state.
Why Max has the same RPM as Pro
Max is designed for higher daily capacity, not higher resale-style throughput. It includes more daily request units and a larger request-count guard while keeping per-minute throughput controlled.
| Plan | Daily request units | Daily request-count guard | RPM |
|---|---|---|---|
| Pro | 1000 | 1000 | 65 |
| Max | 4000 | 4000 | 65 |
If you need higher per-minute throughput for a specific use case, contact Accly before building around that requirement.