Start building with Vitna.
Pick the task in front of you. Each starts with one command and shows what comes back.
Explore the public intelligence API
Needs No key, no signup and no quota.
curl https://vitna.ai/api/v1/markets
{
"classes": [
{ "class": "model", "label": "Language models",
"priceCount": …, "skuCount": …, "providerCount": … },
…
],
"meta": { "generatedAt": "…", "source": "database" }
}Four first requests.
Every command below matches the documentation verbatim, with the response shape each endpoint returns.
curl https://vitna.ai/api/v1/markets
{ "classes": [ { "class": "model", "label": "Language models", "priceCount": …, "skuCount": …, "providerCount": … }, … ], "meta": { "generatedAt": "…", "source": "database" } }- 02The Web API reference
Make a web-data request
Needs A vitna_web_ key. Public signup is closed for now, so keys come by invitation, activated on the Web APIs account page; nothing is billed.
curl https://web.vitna.ai/v1/read \ -H "Authorization: Bearer $VITNA_WEB_KEY" \ -H "Content-Type: application/json" \ -d '{"url":"https://example.com/","format":"markdown"}'{ "url": "https://example.com/", "title": "…", "format": "markdown", "content": "# …" } - 03The Relay guide
Install a model gateway
Needs Node 22.13 or later and your own provider key. Relay asks for no Vitna key.
npm install -g vitna-runtime OPENAI_API_KEY=sk-... vitna-relay # then, in a second terminal curl http://127.0.0.1:8787/healthz
{ "ok": true, "version": "0.2.0" } - 04Anchor's install section
Run local inference
Needs Windows or Linux on x86-64. No account, and the daemon's key is a local placeholder that nothing checks.
# Windows, in PowerShell irm https://github.com/convexityos/vitna-anchor/releases/latest/download/install.ps1 | iex # Linux on x86-64 curl -fsSL https://github.com/convexityos/vitna-anchor/releases/latest/download/install.sh | sh
The installer picks a model that fits the machine, fetches it at a pinned revision with every file checked, and starts the server at http://127.0.0.1:8765/v1 with its chat page. Any OpenAI-compatible client works against it, with the model named as the installer named its directory.
Quickstarts and references
Every product's own documentation.
Each service publishes its own quickstart and full contract, with live stage indicators beside every entry.
Free, no key
PublicPublic intelligence API
Web search API
Private betaSearch
Page reading and extraction APIs
Hosted pilotWeb APIs
Browser rendering API
AlphaUnblocker
Model gateway on your own keys
On npmRelay
Self-run model gateway with policies
On npmRuntime
Local inference engine
Early releaseAnchor
Workload decisions
First workload freeDecision Console
Authentication and key management
Which product issues which credential.
Each product that issues a key issues its own today. One organisation key that Search, the Web APIs and the Unblocker all accept is minted by Vitna during the beta.
No key
Public intelligence API
Nothing to issue. Every read endpoint is public.
The API referencevitna_web_Web APIs
By invitation while public signup is closed. The invitation is activated on the account page, whose owner token then mints, rotates and revokes keys; they last 90 days by default.
Request a Web API keyvitna_search_Search
By invitation, while Search is a private beta.
Request a Search keyvitna_unblock_Unblocker
Issued with alpha access, once a person has checked the fit of your workload.
Request alpha accessvitna_key_Search, Web APIs and Unblocker
One organisation key the three services accept, minted by Vitna during the beta rather than by the console.
One keyYour provider keys
Relay and Runtime
Your provider keys, held on your machine. The Runtime also loads a policy the Decision Console signs.
InstallingNo key
Anchor
Nothing. It runs on your machine and checks no key.
Install AnchorA sign-in
Decision Console
A sign-in, for your organisation's workloads and policies.
Sign in
Local software
Packages
What installs today, from where, under which licence.
- vitna-runtime0.2.0 · Apache-2.0npm
npm install -g vitna-runtime
- vitna-anchorv0.2.1 · Apache-2.0GitHub releases
# Windows, in PowerShell irm https://github.com/convexityos/vitna-anchor/releases/latest/download/install.ps1 | iex # Linux on x86-64 curl -fsSL https://github.com/convexityos/vitna-anchor/releases/latest/download/install.sh | sh
The hosted APIs need no client library: every route is a plain HTTPS request with a bearer key.
Agent tooling
Two hosted MCP servers
Connect agents over Streamable HTTP: 25 tools across Web APIs, research, verification, structural diffs, PDF/Office parsing, dossiers, grids, firehose events, 16 official sources and Google SERP on web.vitna.ai/v1/mcp, plus 10 keyless pricing tools on vitna.ai/api/v1/mcp.
claude mcp add --transport http vitna-web https://web.vitna.ai/v1/mcp \ --header "Authorization: Bearer $VITNA_WEB_KEY"
claude mcp add --transport http vitna https://vitna.ai/api/v1/mcp
Wire Claude Code, Cursor or Codex directly over stateless HTTPS without running a local proxy server.
Deployment, recipes, errors.
Deployment
Recipes
- Verify a figure Take a printed price back to its document and its signed day.
- Cite a figure Quote a price with its date, its document and a lasting URL.
- Extraction recipes Save a page's selectors once and reuse them across a site.
- Profile your traffic Read what your requests consume from the ledger, offline.
Errors and limits
Machine-readable specs and release notes: inspect every OpenAPI, MCP and LLM manifest directly, or read package releases on npm and GitHub.
