agent
Autonomous coding agent
A tool-calling loop that inspects the project, plans, edits and verifies. It stops at approval gates instead of acting silently.
AI coding platformPartially degraded· 1 models enabled
Your Code. Your Flow.
Powered by NordAI.
A terminal agent that reads your project, plans the change and edits files with your approval — backed by a metered AI gateway that routes only to models you are allowed to use.
Free tier · Rp 100.000 for 30 days · unlimited requests · quotas enforced server-side
example session
$ nordai "add request throttling to the public API"
● plan · 4 steps
1 read_file internal/httpapi/router.go
2 search_files "rate limit" · 7 matches
3 edit_file internal/httpapi/middleware.go +48 −6
approval required · apply changes? [y/N]
4 run_command go test ./internal/httpapi/... ✓ pass
✓ done · 18 files read · 2 edited · 1 command · 1 approval
Illustrative transcript. The CLI prints real tool results.
CLI reference →- Metering
- per request, atomic
- Models
- 1 approved
- Sessions
- persistent & resumable
- Audit
- admin actions logged
Capabilities
Built for real repository work
Every capability below exists in the shipped CLI and API. Where something is limited, the limit is shown instead of hidden behind marketing copy.
loop
Multi-step workflows
Plan, tool call, observation, next step. Step budgets and loop detection keep long tasks bounded and cancellable.
gateway
Model routing & fallback
The Go gateway forwards only to models your plan and key allow, with controlled fallback on upstream rate limits.
context
Project-aware context
Repository inspection, file search and context compaction that keeps long sessions inside the model window.
tools
File editing & Git diff
Edits are applied inside the workspace boundary, with a diff you can review and an undo step that restores the previous file.
sessions
Session persistence
Resume a session with /resume, compact it with /compact. Transcripts are stored server-side per account.
security
Permission controls
Manual, auto-edit, read-only and full-access approval modes. Deny rules and command allowlists run before anything executes.
usage
Transparent usage limits
Every request is counted by the API before forwarding. Failed requests are not billed, and resets are shown in WIB.
Architecture
From prompt to verified change
- 01CLI
User prompt
You describe the task in your terminal. The CLI sends it to NordAI with your session context.
- 02Planner
Agent loop
The model plans, selects tools and streams a response. Every step is bounded by a maximum step count.
- 03Workspace
Tool execution
Reads, edits and commands run inside your workspace with approvals, timeouts and path checks.
- 04Gateway
AI router
The Go gateway validates your plan, model allowlist and quota, then forwards to the configured upstream provider.
- 05Result
Verified result
Tests or commands you approved produce the evidence. The agent reports incomplete work instead of claiming success.
This is the implemented flow: the website and CLI authenticate with NordAI credentials, the Go gateway owns entitlements and forwarding, and the upstream key never leaves the API service. No step is simulated.
Model catalogue
Approved models, not a marketing list
Model ids come from the live upstream catalogue. An administrator verifies provider terms before enabling a model for customers.
paid32,000 ctx
NordAI DeepSeek V4.1 Flash
nordai-deepseek-v41-flash
Pricing
One daily pass. Or pay per token.
Prices and limits are read from the NordAI database. The Daily pass is Rp 100.000 for 30 days of unlimited requests. Pay as you go instead bills each successful request from a prepaid balance, with no request quota and no expiry on unused balance. Both can be changed by the operator without a redeploy.
NordAI Pro
UnlimitedRp 100.000/ 30 days
Rp100.000 / bulan · 500 juta token
- 200 requests / day
- 6,000 requests / month
- 20 requests / minute
- 5 concurrent requests
- 16,384 max output tokens
- 5 API keys
- Approved free and paid models
Renews every 30 days. Access activates only after the payment provider confirms settlement to our server.
NordAI Pay as you go
Pay as you goPrepaid/ from Rp 10.000 top-up
Unlimited requests, charged per token
- No request quota and no plan limits — the balance is the only gate
- Every model, billed at its published per-token price
- Charged from your balance only after a request succeeds
- No rate limit
- No concurrency limit
- No output token limit
- Unlimited API keys
- Approved free and paid models
Top up from Rp 10.000. Each successful request is charged from your balance at the published per-token price; a failed request is not charged. Balance has no expiry.
The Daily pass removes NordAI's request quota; pay as you go replaces it with your own balance. Upstream availability still depends on provider capacity, and a per-minute rate and concurrency guard protect the service from runaway clients.
FAQ
Questions developers ask
Still unsure? Open a support request from your dashboard — it is answered by the same people who operate the service.
Contact supportIs usage unlimited?
On the Daily pass, yes: there is no daily or monthly request quota on any model for 24 hours. A per-minute rate and a concurrency guard still apply so one client cannot saturate the upstream provider. The Free plan keeps a fixed allowance. Pay as you go has no request quota either: each successful request is charged from your prepaid balance at the model's published per-token price. Usage is recorded on every plan, but only a limited plan blocks a request.
How does pay as you go work?
You top up a balance from Rp10.000 through a normal QRIS invoice. The balance is credited only when the payment provider confirms settlement to our server. Each successful request is then charged from the balance at the serving model's published input and output price; a failed request is not charged. When the balance reaches zero, new requests are refused until you top up. Unused balance does not expire.
What happens when I reach a limit?
The request is rejected before it reaches the AI router with a structured error that includes your reset time and an upgrade link. On a prepaid plan an empty balance returns a payment-required error with a top-up link instead. Failed requests are not counted as billable usage.
How is a pass activated?
The website creates an invoice on the server, stores the expected amount, and activates the plan only after the payment provider confirms settlement to our server. A browser redirect or a screenshot never activates anything. Because QRIS has no signed callback, the server polls the provider and the invoice page has a Refresh status button.
Can I get a refund?
Access starts as soon as payment is confirmed, so an active 24-hour pass is not refundable. This is shown before you pay. Passes are bought one day at a time; nothing renews automatically.
Which models can I use?
Only models approved by the operator for your plan. New upstream models are disabled by default until their provider terms are verified. The public model page shows what is currently enabled.
Does the CLI hold the upstream key?
No. The CLI authenticates with your NordAI device credential or a customer API key. The upstream provider key lives only in the API service environment and is never sent to a client.
Is my source code stored?
By default the API logs request metadata only — model, tokens, status and latency — never prompts, files or command output. CLI transcripts are stored per account for /resume and can be removed with your account.
Start with the free allowance, upgrade when the work demands it.
Create an account, issue a key, install the CLI. Usage appears in your dashboard the moment the gateway forwards a request.
npm install -g nordai