For solo founders and small software teams
Never forget what your stack is up to.
Relephant connects your cloud, AI APIs, payments and code. It keeps the history, notices what changed and tells you in plain English. Ask for any dashboard and it builds it.
$5 of free credit · No card · No password
You're signed in on this browser.
Open RelephantSince you were here (3 days ago): 3 insights · 412 events
EC2 instance web-2 has been idle for 9 days.
CPU has been under 2% since Sep 30. It costs about $1.10 a day.
AWS · EC2 · 2h ago
Claude API spend is 38% above this point last month.
Most of the increase is on the prod API key.
Anthropic · 5h ago
3 new paying customers this week.
Including your first annual plan.
Stripe · 1d ago
Nine tabs. Zero memory.
Your AWS bill lives in one console. Claude and OpenAI spend live in two more. Revenue is in Stripe, deploys are in Vercel, errors are in Sentry. Each one shows you today. None of them remembers last Tuesday, and none of them talks to the others.
- AWS Billing
- Anthropic Console
- OpenAI Usage
- Stripe
- Vercel
- GitHub
- Sentry
- Supabase
- PostHog
Connect once. It keeps watching.
Connect
Paste a read-only key, or run a one-step AWS setup. Each connector lists exactly which permissions it needs and why.
It watches
Relephant syncs each source on a schedule, keeps the history and flags what changed: a stopped server, a spend spike, a failed payment.
You get the relevant bits
A short feed of insights written by Claude, each linked to the events behind it. Ask a question or ask for a dashboard in plain English.
A feed that reads like a teammate's notes.
Not 400 alerts. A handful of sentences about what changed, why it matters and what it costs, each linked to the events behind it. Switch to Mixed or Raw whenever you want the receipts.
- Insights only run when something changed, so quiet weeks cost almost nothing.
- Raw events are always free and kept for 90 days.
- Mark an insight as not useful and Relephant learns what you don't care about.
4 failed payments since Monday, twice the usual.
Three were declined cards on the Pro plan. Stripe will retry two of them tomorrow.
Stripe · 3h ago
Based on 4 events
- Mon 09:14stripe.invoice.payment_failedin_1Q8…a2$49.00
- Tue 16:40stripe.invoice.payment_failedin_1Q9…f7$49.00
- Wed 11:02stripe.invoice.payment_failedin_1QA…0c$199.00
- Thu 08:55stripe.invoice.payment_failedin_1QB…9e$49.00
14 AWS events · 2–3 PM
Your whole stack on one map.
Every connected service, its health and its spend at a glance. Hover for details. Click AWS to open up Lambda, EC2, S3 and the rest.
AWS · production
Healthy · synced 4 min ago
MTD $41.20 · 7 running resources
Open dashboard
Ask for a dashboard. Get a dashboard.
Describe what you want to see. Relephant picks the data, lays it out and keeps it up to date. Refreshing is free: it re-runs the queries, not the AI. Change it by asking again.
- What are all my costs by month? Add a month selector.
- What are my users doing?
- What infrastructure am I paying for but not using?
- How far into the funnel do my users get?
What are all my costs by month? Add a month selector.Build
- Reading AWS costs
- Reading Claude and OpenAI spend
- Laying out 4 widgetsRefreshes every 6 hours · free
Total spend, September
$1,284.50
12% vs August
Claude API, September
$498.00
19% vs August
By source, September
Ask about any chart.
Why did costs jump in March? The answer appears on the chart itself, pinned to the bar it's about. Keep the notes you want, and everyone you share the dashboard with sees them.
Why did costs jump in March?
Monthly cloud spend
A NAT gateway created on Mar 3 in us-east-1 accounts for $212 of the $260 increase. It has carried almost no traffic since Mar 20.
Keep note · Dismiss
Works with the tools you already pay for.
Fourteen sources at launch. Read-only wherever the service allows it.
- AWSCosts by service, EC2, Lambda, S3, RDS and more
- Google CloudCosts from your billing export, plus Compute Engine and Cloud Run
- Gemini APIRequests, tokens and estimated spend
- Google WorkspaceUsers, admin changes and sign-in activity
- AnthropicClaude API spend and tokens by model and key
- OpenAISpend and tokens by model and project
- StripeMRR, revenue, failed payments and disputes
- GitHubCommits, pull requests and Actions runs
- VercelDeployments, failed builds and usage
- CloudflareTraffic, cache hit rate and Workers
- SupabaseProjects, service health and usage
- SentryNew issues and error volume
- PostHogActive users, top events and funnels
- Custom eventsSend any event from your code with one HTTP call
- Missing one?Tell us
Logos are trademarks of their owners. Showing them doesn't mean they endorse Relephant.
Read-only by design.
Read-only access
We ask for the narrowest access each service offers and list every permission before you connect. Where a service has no read-only key (Vercel tokens, the Anthropic and OpenAI admin keys), we say so and only call read endpoints.
Encrypted keys
Credentials are encrypted with AWS KMS, decrypted only in memory while we fetch your data, and never shown again. The app shows only a hint like rk_live_…9f2a.
Keys never reach the AI
Claude sees summaries of your data, never credentials. Email addresses are masked before anything is sent.
Delete means delete
Remove a connection and its key is gone at once and its data within a day. Delete a project and everything in it is erased within 30 days.
What each connector uses
| Service | What you give us |
|---|---|
| AWS | A read-only IAM role you create with our CloudFormation template, protected by an external ID. No access keys. |
| Google Cloud, Gemini API | Keyless access from our AWS account, set up with a script you run in Cloud Shell. A service-account key works too. |
| Google Workspace | Domain-wide delegation with read-only admin scopes. |
| Anthropic | An Admin API key. We only call the usage and cost endpoints. |
| OpenAI | An admin key. We recommend a read-only one. |
| Stripe | A restricted key with read access. Secret keys are refused. |
| GitHub | A fine-grained token with read permissions. |
| Vercel | A team token. Vercel has no read-only tokens, so we only call read endpoints. |
| Cloudflare | An account token with read permissions. |
| Supabase | A read-only personal access token for the Management API. |
| Sentry | An organization integration token, or a personal token. |
| PostHog | A personal API key for US or EU Cloud. At most 60 queries an hour. |
| Custom events | Nothing. You send events to a private endpoint with a token we give you. |
Pay for what it thinks. Nothing else.
Start with $5 of free credit. No card needed.
- $5 free
- To start. One per person, while the day's allowance lasts.
- $10, $25, $50 or $100
- Top up when you like. Credit never expires.
- Cost + 25%
- What Anthropic charges us for the tokens, plus 25%.
You stay in control
- An email when your balance drops below $1.
- At $0, AI work pauses. Syncing and dashboard refreshes keep going.
- Insights check for changes every 6 hours by default (or hourly, daily, or never) and only use the AI when something changed. At most $0.50 a day per project.
- People you share with can ask questions only if you allow it, up to a monthly limit you set.
| What you do | Typical cost |
|---|---|
| Build a dashboard from a prompt | about $0.16–$0.65 |
| Change a dashboard by prompt | about $0.06–$0.39 |
| Ask a question about a dashboard | about $0.06–$0.32 |
| Write feed insights (one run) | about $0.06–$0.31 |
| Connect, sync, view, refresh and share | Free |
Estimates for Claude Opus 5.5. Bigger projects use more tokens. You see an estimate before every action and the exact cost after.
How the price works
We pay Anthropic for every token Claude uses on your behalf. You pay that cost plus 25%. A question that uses $0.12 of tokens costs you $0.15. Every charge in your ledger shows the model, the tokens, Anthropic's price and our 25%.
Questions, answered.
What is Relephant?
One place that watches your project's cloud, AI APIs, payments and code. It keeps the history, notices what changed and tells you in plain English. It also builds dashboards when you ask and answers questions about them.
What can I connect?
AWS, Google Cloud, the Gemini API, Google Workspace, Anthropic, OpenAI, Stripe, GitHub, Vercel, Cloudflare, Supabase, Sentry and PostHog. You can also send your own events from your code to a private endpoint. Missing something? Email us.
Is it safe to give you my keys?
We ask for the narrowest access each service offers and show every permission before you connect. AWS uses a read-only role you create with our template, not access keys. Keys are encrypted with AWS KMS, never shown again and never sent to the AI. Remove a connection and its key is deleted at once.
What does the AI see?
Only what an answer needs: metric totals, recent events, resource names and the data on the dashboard you're asking about. Never credentials. Email addresses are replaced with anonymous tokens before anything is sent. We use Claude through Anthropic's API.
How does pricing work?
You prepay for credit. New accounts get $5 of free credit to start (one per person, while the day's allowance lasts). AI work, meaning building or changing a dashboard, answering a question or writing feed insights, costs what Anthropic charges us for the tokens plus 25%. Everything else is free. Your ledger shows every charge with the model, the tokens and the price.
Why cost plus 25%?
It pays for the servers, the syncing and our time, and it keeps the incentive simple: we earn when Relephant does useful work for you, not when you forget to cancel a plan.
What happens when I run out of credit?
Your data keeps syncing and your dashboards keep refreshing. New insights, questions and dashboard changes pause until you top up. If a request is already running when you reach $0, we let it finish, and the few cents it costs come off your next top-up.
Do credits expire? Can I get a refund?
Credit never expires. If you want a refund of credit you bought and haven't used, email hello@relephant.ai.
Can people I share with run up my bill?
Only if you let them ask questions or edit by prompt, and only up to the monthly limit you set for that share ($5 unless you change it). Viewing is always free.
Which AI model do you use?
Claude by Anthropic, currently Claude Opus 5.5. If we change models, prices follow the actual token cost automatically, and your ledger shows which model did each job.
How fresh is the data?
Each source syncs on its own schedule, from every 15 minutes for Stripe events to every 6 hours for AWS costs. Billing data arrives when the provider publishes it, which for cloud costs can be several hours late.
Does connecting AWS cost me anything?
A little. AWS charges $0.01 per Cost Explorer request, which adds about $1–2.50 a month to your AWS bill. You can turn cost syncing off for that connection.
Can I use it for several products, or with a team?
Yes. Create a project for each product. Share any dashboard or the whole overview by email, and choose whether each person can edit or ask questions.
How do I delete my data?
Remove a connection and its key is deleted at once and its data within a day. Delete a project and everything in it is erased within 30 days. To close your account, email hello@relephant.ai.
Why an elephant?
An elephant never forgets, and Relephant only shows you what's relevant. We couldn't resist.
Give your stack a memory.
$5 of free credit. No card. Your first project is ready the moment you sign in.
By signing in you agree to the Terms and Privacy Policy.