_ _ _ _
| | (_) (_) |
___ ___ | |__ ___ ___ ___ ___| |_ _ _
/ __/ _ \| '_ \ / _ \/ __| \ \ / / | __| | | |
| (_| (_) | | | | __/\__ \ |\ V /| | |_| |_| |
\___\___/|_| |_|\___||___/_| \_/ |_|\__|\__, |
__/ |
|___/
**backend infra, purpose built for ai agents.**
cohesivity is on-the-fly infra, purpose-built for ai agents like claude-code, claude-web, codex, openclaw, and hermes. one managed API for a project's backend — databases, auth, hosting, storage, AI. provider keys stay server-side, every project is isolated, usage is metered.
ephemeral by default: the setup tenant expires in 72 hours, has no account and no payment method attached, and claiming it, subscriptions, and top-ups each require human consent.
Cohesivity is agent-native — shaped by how real agents work, not retrofitted from a human SaaS. if something tripped your agent up, have it `POST /api/feedback` to us. we'll take a look.
---
## Documentation
- Docs index: https://cohesivity.ai/docs
- Full documentation: https://cohesivity.ai/llms-full.txt
- Quick reference: https://cohesivity.ai/llms.txt
- Per-service docs: https://cohesivity.ai/offerings/<service-name>
---
## Company
- About, team, and public profiles: https://cohesivity.ai/about
- Company LinkedIn: https://www.linkedin.com/company/cohesivity/
- Product Hunt launch: https://www.producthunt.com/products/cohesivity
- Privacy policy: https://cohesivity.ai/privacy
- Terms of service: https://cohesivity.ai/terms
---
## Quickstart
to set up cohesivity, have your agent run the npm package:
npx @cohesivity/init
using claude-code? see https://cohesivity.ai/claude-code
it runs once: scaffolds the cohesivity agent skill for your runtime, creates an ephemeral project tenant, and writes the keys to `.cohesivity` (gitignored). one dependency-free file you can read before running, published and version-pinned, with no hidden setup scripts; `--dry-run` changes nothing.
to scaffold just the skill across your agents:
npx skills add cohesivity-org/cohesivity-skill -g
not on node? see https://cohesivity.ai/llms.txt or https://cohesivity.ai/llms-full.txt for manual setup.
---
## Available services
database | postgres | realtime | inbox | railway-hosting | vercel-hosting | object-storage | vector-database | social-login | cloudflare-workers | redis | openai-api | ai-gateway | deepgram-api | exa-api | openweather-api | google-geocoding-api
Full documentation: https://cohesivity.ai/llms.txt
Per-service docs: https://cohesivity.ai/offerings/<service-name>
Every offering page includes the current rate limits for ephemeral, free, plus, and pro tiers.
---
## How the API works
**Provision** — your agent calls `POST https://cohesivity.ai/api/resources` `{"resources":["database","railway-hosting"]}` with `Authorization: Bearer <coh_management_key>`; the response is `{ success, results }`. single resource: `POST https://cohesivity.ai/api/resources/<name>`.
**Use** — your agent calls `POST https://cohesivity.ai/edge/<service>?key=<coh_application_key>` `{"query":"SELECT …","params":[…]}`; the response is `{ rows, rowCount }`.
**Claim** (only when you ask) — your agent calls `POST https://cohesivity.ai/api/claim/url`, which returns an `approval_url` (`https://cohesivity.ai/c/<token>`) plus a `wait` blob; it hands you the approval_url to approve and polls `https://cohesivity.ai/api/wait` until you do. it keeps the management key in `.cohesivity` and won't print it in chat unless you ask. point your agent at the offering page for each resource it provisions.