For anyone building apps that use AI

Keep all yourAI keys secured.

knectar is one safe place for your OpenAI, Anthropic, and other AI keys. Each app gets its own knectar key instead. Just simply switch off the knectar key when it leaks, and nothing else breaks. You also get a backup model of your choice when a provider goes down, and every request shows what it cost.

Join the waitlist See how it works

  1. Your app

    Today your OpenAI key sits in every app's .env. One leak means pasting a new key into all of them. With knectar, the app holds only its knectar key, in the same OPENAI_API_KEY spot.

    sk-bf-a91f

  2. knectar

    Checks the knectar key, adds your OpenAI key and sends the request on. Your app never sees your OpenAI key.

    OpenAI key added

  3. OpenAI

    Answers. The reply comes back through knectar to your app, and the request shows up in your logs with what it cost.

    openai/gpt-4o-mini

  4. Your backup model

    If OpenAI fails or times out, the same request goes to the backup model you picked. You write no retry code.

    anthropic/claude-haiku-4-5

  5. Saved answers

    Turn them on with one header. A repeat request gets the saved answer and never reaches OpenAI, so OpenAI never bills it. Best for repeat jobs, not chat.

    x-bf-cache-key: my-job

knectar is running now. Our server measured these with our own test key in the last ten minutes. Your apps will use the AI keys you connect, not ours.

Status

measuring

Models our test key can reach

measuring

AI providers on our test key

measuring

One test request

measuring

Your dashboard

Logs: one row for every AI request.

Each row shows the model, how long it took, the tokens and what it cost. Click a row to read what was sent and what came back. These are sample rows.

  1. 01 One row for every AI request
  2. 02 Click a row to read the prompt and the reply
  3. 03 Filter by key, model or status
Search ⌘K Docs

LLM logs

Observability

Requests
1,284
Success
99.5%
Avg latency
852 ms
Tokens
1.86M
Cost
$1.996

Request volume last 24 hours

15:00 14:00 UTC
  • success
  • error
  • Peak 103

Top models

all time

  • openai

  • anthropic

  • openai

Usage by key

all time
StatusObjectModelKey
14:55:24processingchat.completionclaude-haiku-4-5anthropicbilling-app8121.24 s$0.00080
14:55:17processingchat.completiongpt-4o-miniopenaisupport-bot156380 ms$0
14:55:01successchat.completionclaude-haiku-4-5anthropicbilling-app550642 ms$0.00110
14:54:39successchat.completiongpt-4o-miniopenaisupport-bot12038 ms$0
14:54:16errorchat.completionclaude-haiku-4-5anthropicbilling-app0214 ms$0
14:53:53successembeddingtext-embedding-3-smallopenaidocs-search1,024188 ms$0
14:53:27successchat.completiongpt-4o-miniopenaibilling-app1,792910 ms$0.00050
14:53:00successchat.completionclaude-haiku-4-5anthropicbilling-app3,9842.21 s$0.00740
14:52:32successchat.completiongpt-4o-miniopenaisupport-bot17642 ms$0
14:52:05successchat.completiongpt-4o-miniopenaisupport-bot360730 ms$0.00010
14:51:28errorchat.completiongpt-4o-miniopenaidocs-search6202.40 s$0
14:50:57successembeddingtext-embedding-3-smallopenaidocs-search512196 ms$0
14:50:23successchat.completionclaude-haiku-4-5anthropicbilling-app476588 ms$0.00100
14:49:46successchat.completionclaude-haiku-4-5anthropicsupport-bot2,8501.76 s$0.00530
14:49:00successchat.completiongpt-4o-miniopenaibilling-app10251 ms$0
14:48:17successchat.completiongpt-4o-miniopenaidocs-search1,200820 ms$0.00030
14:47:28errorchat.completiongpt-4o-miniopenaibilling-app0180 ms$0
14:46:28successchat.completionclaude-haiku-4-5anthropicbilling-app2,2601.98 s$0.00440
14:45:16successembeddingtext-embedding-3-smallopenaidocs-search768204 ms$0
14:43:48successchat.completiongpt-4o-miniopenaisupport-bot25457 ms$0
14:41:38successchat.completiongpt-4o-miniopenaibilling-app538664 ms$0.00010
14:38:38successchat.completionclaude-haiku-4-5anthropicsupport-bot3,5001.12 s$0.00710
14:31:18successchat.completionclaude-haiku-4-5anthropicdocs-search1,5521.49 s$0.00300
14:11:18successchat.completiongpt-4o-miniopenaibilling-app76470 ms$0
1-24 of 1,284 (page 1 of 52)
  1. Step 1 One row for every AI request
  2. Step 2 Click a row to read the prompt and the reply
  3. Step 3 Filter by key, model or status

Your dashboard

knectar keys: one for each app.

Create a key, swap it for a new one or delete it. The dashboard says mint for create, rotate for swap and virtual key for a knectar key. You see the full key once, so copy it then. Try it here. Nothing you type leaves your browser.

  1. 01 Create one key for each app
  2. 02 Swap a key for a new one
  3. 03 Delete a key and only that app stops
Search ⌘K Docs

Virtual keys

Governance

  • billing-app Active

    948a••••••••••••••••••••••••••••26a9

    $61.40 of $250.00 monthly600 requests every minute2 providers

  • support-bot Active

    5a56••••••••••••••••••••••••••••8093

    $12.80 of $50.00 monthly120 requests every minute1 provider

  • docs-search Active

    0a82••••••••••••••••••••••••••••857d

    No limits

  1. Step 1 Create one key for each app
  2. Step 2 Swap a key for a new one
  3. Step 3 Delete a key and only that app stops

Questions people ask before connecting an app.

No. Three lines change: the address your app sends AI requests to, the key and the model name. The model name starts with its AI provider, like openai/gpt-4o-mini. The rest of your code stays the same.

Get your knectar key when sign-ups open.

Leave your email and we will write to you once, when sign-ups open. Nothing to pay today, and nothing to install. We run knectar for you.

your knectar key, ready when sign-ups open
AI providers you use today

One email when sign-ups open. Nothing else.

How we handle your email