No. Stellar Intel never holds user funds, user keys, fiat, or KYC data. You sign
every transaction in your own wallet; the anchor takes custody under SEP-24/SEP-6;
Stellar enforces atomicity. See docs/NON_CUSTODY.md.
The anchor — not Stellar Intel — is the settlement party. We surface that risk
before you choose: each anchor carries a reputation score built from real
outcomes (fill rate, slippage, settle latency). A failing anchor ranks down and
its failures are recorded. If an outcome is wrong, you can file a dispute, which
resolves on signed, replayable evidence. See
docs/ANCHOR_REPUTATION.md.
We compare live rates across every integrated anchor and rank by net landed value (rate − fees − slippage − historical fill-rate penalty), not headline rate — then let you execute in one click. Plus a public reputation oracle and an MCP agent surface. It's the execution layer, not just a price page.
Most often: the anchor doesn't serve that corridor, or it only exposes a SEP the
rate engine doesn't yet quote. Historically SEP-6-only anchors (e.g. Cowrie) were
dropped because the flow was SEP-24-only — SEP-6 support is being added (see
docs/SEP_COMPLIANCE.md). Failed anchors render as
"unavailable" rather than disappearing silently.
See the registry in constants/anchors.ts and the live
app. Coverage expands via anchor onboarding.
Yes — an MCP server exposes pricing/comparison and a user-signed execute path. The
agent cannot move funds on its own; every executing call must be signed by the
user's wallet. See docs/MCP.md.
A typed client (@stellarintel/sdk) is on the roadmap (v4). Today you can call the
HTTP API directly. See docs/SDK.md.
Read CONTRIBUTING.md, pick a
good-first-issue,
and open a PR. The docs/COOKBOOK.md has end-to-end recipes.