Starter
Free
Implement self-healing now
No credit card required
What's included
- API request self-healing
- Access to your self-healing data
So your apps, workflows and agents stay up.
Manifest is a resilience layer that prevents downtime due to bad API requests.
3 lines of code, available for NodeJS and
Python. It only intercepts API client errors, not the successful requests.
import { manifest } from 'manifest'; manifest(); // Once, at startup. // Keep making your API calls as usual.
Bad API requests are fixed on the fly and sent back, preventing downtime even if the error hasn't been seen yet
| Date | Outcome | HTTP status | Service | Related issue |
|---|---|---|---|---|
| 15:36 | Healed | 400 | api.stripe.com | Received unknown parameter: metadata_notes |
| 11:29 | Healed | 400 | openai | Unsupported parameter: 'max_tokens' is not supported with this model |
| 08:50 | Healed | 400 | anthropic | Extra inputs are not permitted: system.0.cache_control |
| 07:11 | Healed | 404 | openai | The model 'gpt-4-32k' does not exist |
| 05:51 | Healed | 422 | api.internal | Unknown field: customer_notes |
| 04:52 | Healed | 400 | api.stripe.com | This value must be greater than or equal to 50: amount |
| 03:24 | Healed | 400 | api.stripe.com | Received unknown parameter: metadata_notes |
| 02:48 | Healed | 400 | anthropic | Extra inputs are not permitted: system.0.cache_control |
So you find a proper fix later. In the meantime all requests are patched.
Request has been healed by Manifest
How to stop producing this failure
Copy for agentTo fix it at the source, send the value as metadata[notes] instead of metadata_notes
Observe and improve your app to make it more resilient
HTTP 422
{ "param": "scheduled_at", "message": "Input should be a valid datetime, expected 2026-09-08T18:30Z" }
HTTP 422
{ "param": "scheduled_at", "message": "Input should be a valid datetime, expected 2026-09-08T18:30Z" }
"template": "push-day", "scheduled_at": "08/09/2026 18:30" "scheduled_at": "2026-09-08T18:30Z"
HTTP 200
{ "error": null, "statusCode": 200 }
Stop letting bad API requests take your app down.
Start for freeManifest is integrated into apps, workflows and agents to make them more resilient. As systems become less deterministic and more connected, chances are that connections eventually break, causing downtime that affects production and user experience.
Each connector will eventually break some day
One failed request breaks the whole process
Depending on legacy or poorly-versioned APIs
API calls built at runtime are more likely to fail
Manifest fixes errors that cannot be caught at build time.
The vendor sent a warning email a few months ago to someone on the team, but it got lost and no one read it.
POST /v1/orders Api-Version: 2026-07-29 { "shippingMethod": "express" "shipping_method": "express" }
An agent built the request at runtime and asked for one day more than the endpoint allowed.
POST /invoices/export { "account_id": "acc_2c7f", "from": "2026-08-01", "to": "2026-09-01", "to": "2026-08-31", "format": "csv" }
Resulting in an empty field sent to the API. Not caught by any code reviewer
POST /coach/messages { "thread_id": "thr_8f2a", "user_id": "", "user_id": "Yzcc…HnL2", "text": "Done for today", "sent_at": 1788633727 }
Free
Implement self-healing now
No credit card required
What's included
Custom pricing
For teams with larger or more specific needs.
What's included
Don't wait until your production breaks. Start with Manifest now and turn failed requests into successful ones.