Kaval
FreeNot checkedFreshness gate for AI agents: verify a belief is still true against the live world before you act.
About
Freshness gate for AI agents: verify a belief is still true against the live world before you act.
README
Open-source client libraries for Kaval — the freshness gate for AI agents. Before your agent acts on a belief it already holds (a cached fact, a stored field, a retrieved RAG chunk, a prior answer), Kaval independently re-derives the truth from the live world and returns a typed, abstaining verdict your agent can branch on.
These are thin HTTP clients for the hosted Kaval API (https://api.usekaval.com). Create an API
key at usekaval.com.
The Node, Python, and MCP surfaces automatically attach a unique idempotency key to billable operations and reuse it for one bounded retry only when the transport outcome is ambiguous or the API is still finalizing that operation.
| Package | Language | Install | Source |
|---|---|---|---|
| @usekaval/kaval | Node / TypeScript | npm i @usekaval/kaval |
sdks/node |
| kaval | Python | pip install kaval |
sdks/python |
| @usekaval/mcp | MCP server | npx -y @usekaval/mcp |
packages/mcp |
Node
import { Kaval } from "@usekaval/kaval";
const kaval = new Kaval({ apiKey: process.env.KAVAL_API_KEY });
const proof = await kaval.audit({
text: "Acme is eligible for a $12,000 refund",
as_of: new Date().toISOString(),
intended_action: "Issue the refund",
materiality: "critical",
reversibility: "irreversible",
});
const gate = await kaval.gateAction({
proof_id: proof.proof_id,
material_claim_ids: proof.action_decision.material_claim_ids,
threshold: proof.action_decision.threshold,
action: proof.research_contract.action,
});
if (
gate.enforcement?.controlApplied === true &&
gate.enforcement.executionAllowed !== true
) {
throw new Error("Kaval blocked the action");
}
if (
gate.enforcement === undefined &&
(gate.state !== "current" || gate.decision.decision !== "ALLOW")
) {
throw new Error("Kaval did not allow the action");
}
// controlApplied === false is shadow mode: observe wouldAllow without controlling the action.
const { act, status, reason } = await kaval.verify({
belief: "Acme is on our Enterprise plan",
url: "https://billing.acme.com/account",
held_at: "2026-03-01T00:00:00Z",
});
if (!act) {
// status ∈ stale | contradicted | unsupported | insufficient — re-research before acting.
}
Python
from kaval import KavalClient
kaval = KavalClient(api_key=os.environ["KAVAL_API_KEY"])
decision = kaval.verify(belief="Acme is on our Enterprise plan")
if not decision["act"]:
... # re-research before acting
MCP
KAVAL_API_KEY=kv_live_… npx -y @usekaval/mcp
Exposes proof_audit + proof_gate for the full action-verification protocol, plus
currentness_verify, currentness_check, …_extract_and_check, …_scan_store, …_monitor, and
report_outcome over stdio. See packages/mcp.
API origin env vars
Two names exist on purpose — they are not interchangeable:
| Consumer | Variable | Reads env? |
|---|---|---|
| Python SDK / MCP | KAVAL_BASE_URL |
yes |
Node @usekaval/kaval |
— | pass baseUrl in constructor |
Marketing site proxy (apps/web) |
KAVAL_API_URL |
yes (server only) |
Use the same origin value in both vars when self-hosting (e.g. http://localhost:8787). See
SELFHOST.md in the core repo.
Development
pnpm install
pnpm check # build + lint + typecheck + test (the JS packages)
# Python SDK
cd sdks/python && pip install -e ".[dev]" && pytest
License
Installing Kaval
This server has no published package — it is built from source. Open the repository and follow its README.
▸ github.com/LufeMC/kaval-clientsFAQ
Is Kaval MCP free?
Yes, Kaval MCP is free — one-click install via Unyly at no cost.
Does Kaval need an API key?
No, Kaval runs without API keys or environment variables.
Is Kaval hosted or self-hosted?
A hosted option is available: Unyly runs the server in the cloud, no local setup required.
How do I install Kaval in Claude Desktop, Claude Code or Cursor?
Open Kaval on unyly.org, pick your client tab (Claude Desktop, Claude Code, Cursor) and press Install — the config is generated automatically, no JSON editing.
Related MCPs
Fetch
Web content fetching and conversion for efficient LLM usage.
AWS KB Retrieval
Retrieval from AWS Knowledge Base using Bedrock Agent Runtime.
by modelcontextprotocolSpring AI MCP Server
Provides auto-configuration for setting up an MCP server in Spring Boot applications.
llm-analysis-assistant
A very streamlined mcp client that supports calling and monitoring stdio/sse/streamableHttp, and can also view request responses through the /logs page. It also
by xuzexin-hzCompare Kaval with
Not sure what to pick?
Find your stack in 60 seconds
Author?
Embed badge for your README
Browse similar
All ai MCPs
