SolRouter

Identity

SolRouter describes itself as a cryptographically private inference layer for Solana, combining Arcium's MPC network, an AWS-hosted TEE, and OpenServ's reasoning stack. Public materials describe a devnet beta. Because SolRouter stacks two distinct confidentiality mechanisms (MPC and TEE), this record tags both — but neither has an independent attestation or architecture review behind it yet in this seed batch.

API

Privacy

Scale

On-chain

Change log

• 2026-08-24 — record: seeded from provider roster v1.0 (Unverified) (source)
• 2026-09-14 — privacy.mechanism: X claim (e_privacy): People keep asking how we're different from other private AI companies. Most stop at one piece: encrypted transport, or a trusted enclave. So far we haven't seen anyone else shipping all three: client-side encryption, an attested enclave, and open weights on their own nodes. (Claimed) (source)
• 2026-09-21 — api.models: X claim (a_launch): Trading automations and perps are now live on Solrouter Web Type what you want: - Price triggers - Daily DCA - Phoenix perps + stop losses Everything settles inside the enclave. Keys never leave the TEE. Fills + alerts land in Telegram. Live now: https://www.solrouter.com/ (Claimed) (source)
• 2026-10-05 — api.models: X claim (a_launch): 152 PRs Later In September we moved from browser to mobile, bringing users a significantly more capable app, with new models, live trading, and better UX. ... The new models We don't rent frontier capacity and resell it. Every model we serve runs inside an enclave we operate, on GPUs we control... added two bigger models; - Qwen 3-Coder 30B - Qwen 3.5 122B ... Products in development We’ve finally reached a level of product maturity and composability that allows us to build and scale products much faster. ... - Router Labs, our B2B arm... - A local proxy for coding tools... - A native macOS app... October outlook - Finish development on Plugin and Desktop - Close our first paying institutional clients - Complete testing and get the app into the App Store and Google Play - Announce our ecosystem partners we've been cooking with behind closed doors (Claimed) (source)
• 2026-10-05 — scale.tokens_per_day: X claim (c_scale): Spent today optimizing memory, context & latency → Replies now stream as they’re written instead of arriving as one blob → Long-chat time-to-first-word: 2.1s → 0.67s → Narrower context selection: ⅓ of the tokens per message → Targeted memory extraction, meaning lighter workloads & lower latency (Claimed) (source)

The change log is append-only: corrections are recorded as new entries, never rewrites. Full feed via /api/v1/changes.

Sources

1. [roster] docs/decentralized-inference-provider-roster.md (2026-08-24)
2. [website] https://solrouter.com (2026-08-24)
3. [x-account] https://x.com/SolRouterAI (2026-08-24)