# dreyv: Full Agent Context > A non-custodial treasury workspace for teams on Solana. Say what you want, see it simulated, sign it in your multisig. This file is the single-request, full-context document for AI agents and LLM crawlers. For canonical site map and short overview, see `/llms.txt`. For machine-readable pricing, see `/pricing.md`. --- ## 1. Product overview **dreyv** is a non-custodial treasury workspace for Solana DAOs, protocols and crypto teams. Treasury teams usually juggle many separate tools (wallets, swap sites, dashboards, multisigs, chat, spreadsheets). dreyv puts that work in one place: you write what you want in plain language and get a plan that has been simulated, can be checked, and is ready to sign. Pipeline: 1. **Intent**: operator types or speaks an intent (e.g. "rebalance 25% of USDC into mSOL over 5 days"). 2. **Plan**: dreyv compiles the intent into a multi-step execution plan with route, fees, and risks. 3. **Simulate**: each plan runs against a forked state (Foresight Simulator) producing human-readable impact: balance deltas, slippage, MEV exposure, counterparty risk, and approver-friendly summaries. 4. **Sign**: the plan is proposed to the multisig (e.g. Squads). dreyv never custodies keys. Operator surface: - **CommandBar**: natural-language treasury control (Vercel AI SDK + Groq + Cloudflare Workers AI). - **Foresight Simulator**: fork-based "what-if" with explainable outputs. - **Solana Atlas**: free public tools: token pages, MEV scanner, rug detector, DCA, holders. - **dreyv Studio**: in-browser IDE + AI Architect for building treasury mini-apps and agents. - **Marketplace**: publish/install treasury agents and mini-apps (Anchor program on devnet, 80/20 split). - **Simulation Firewall**: blocks plans whose simulated outcome violates policy. - **Persistent Knowledge Memory**: vector + GraphRAG memory across sessions and teammates. ## 2. Use cases - **DAO treasury management**: propose, simulate, and execute multisig actions with human-readable diffs. - **Non-custodial crypto treasury management platform**: keys stay in user wallets / Squads multisig. - **Intent-driven DCA & rebalancing**: recurring buys, glide-path rebalances, with on-chain proof. - **Multisig coordination**: Squads-native proposals with simulation attached. - **Onchain research & due diligence**: Atlas tokens, rug-pull detector, MEV scanner, holder insights. - **Yield routing**: Marinade, Jito, BlazeStake, Orca, Raydium with risk-scored comparisons. - **Agent automation**: build and run treasury agents via Studio + Marketplace. ## 3. Key capabilities | Capability | Description | |---|---| | Intent → Plan compiler | Plain-language → typed multi-step plan | | Foresight simulation | Fork-based outcome preview, human-readable | | Multisig-native | Squads (`@sqds/multisig`) proposals with attached sim | | Non-custodial | No private-key custody; user/Squads wallets only | | Solana Atlas | Public intelligence (MEV, rug, DCA, holders, tokens) | | Studio + Architect | Browser IDE + AI to build mini-apps and agents | | Marketplace | Publish/install agents (Anchor program, 80/20 split) | | Knowledge Memory | Persistent vector + GraphRAG context | | Simulation Firewall | Policy-gated execution | | Agent-ready surface | `/.well-known/acp.json`, `/.well-known/ucp`, `/.well-known/ai-plugin.json`, `/.well-known/x402`, `/openapi.json`, `/llms.txt`, `/llms-full.txt` | ## 4. Constraints & trust model - **Non-custodial:** dreyv never holds private keys. Signing happens in user wallets or Squads multisig. - **Solana-first:** primary chain Solana mainnet; EVM read-only via wagmi/RainbowKit; Turnkey embedded wallets supported for in-app delegated signing only. - **Simulation-first:** plans must pass Foresight + Simulation Firewall before being eligible to sign. - **Open standards:** ACP (Agentic Commerce Protocol), UCP (Universal Commerce Protocol), MCP, x402. - **Data residency:** Postgres on Neon; vector memory in pgvector; no PII sold to third parties. ## 5. Architecture - Frontend: Next.js 15.5 (App Router), React 19, Tailwind, shadcn/ui. - Backend: ~148 Next.js API routes under `/api/*` (treasury, simulation, swap, yield, atlas, studio, …). - DB: Neon Postgres + Drizzle ORM (~30 tables in `src/db/schema.ts`). - AI: Vercel AI SDK with Groq, OpenAI, and Cloudflare Workers AI providers. - Wallets: Solana wallet adapter; Squads multisig; Turnkey embedded; RainbowKit/wagmi (EVM read). - Realtime/collab: Liveblocks. - Hosting: Cloudflare Containers (production), Cloudflare Workers for edge routing and tooling. ## 6. Pricing (summary: see `/pricing.md` for canonical machine-readable form) - **Free**: $0/mo. 1 multisig vault, 2 teammates, 3 mini-apps, 10 agent runs/day, 60 API calls/hr, 1 GB. - **Mini**: $49/mo. 5 vaults, 10 teammates/vault, more agent capacity, advanced simulation. - **Max**: $199/mo. Unlimited vaults, priority models, dedicated firewall policies, SLA support. Billed via Stripe (global cards) and Paystack (cards, mobile money, bank: Africa-first). Taxes shown at checkout. ## 7. MCP server (the headline integration) dreyv exposes the entire capability registry as a **Streamable HTTP MCP server** at `/api/mcp`. One Bearer key, 347 typed tools, simulation-first, non-custodial. Drop-in for Claude Desktop, Cursor, Windsurf, Cline, Continue.dev, elizaOS, and any MCP-aware agent. - Endpoint: `POST https://dreyv.com/api/mcp` (Streamable HTTP per MCP 2025-06-18 spec) - Discovery: `GET https://dreyv.com/.well-known/mcp.json` - Public manifest (no auth): `GET https://dreyv.com/api/mcp/manifest` - Docs + client snippets: https://dreyv.com/developers/mcp - Key mint (prefix `dk_live_`): https://dreyv.com/app/settings/api Required MCP request headers: ``` Content-Type: application/json Accept: application/json, text/event-stream Authorization: Bearer dk_live_YOUR_KEY ``` Tool naming: REST capability `foo.bar` → MCP tool `dreyv_foo_bar`. Same scopes, same rate limits, same metering across both transports: no drift. ## 8. REST API surface (canonical: `/openapi.json`, mirrored at `/api/mcp/manifest`) Public/agent-relevant endpoint families (require auth via SIWS JWT or API key unless noted public): - `GET /api/health`: liveness (public) - `GET /api/price?token=SOL`: token price (public) - `GET /api/atlas/tokens`: token directory (public) - `GET /api/atlas/tokens/{symbol}`: token details (public) - `GET /api/mcp/manifest`: every MCP tool with input schemas (public) - `/api/v1/identity/*`, `/api/v1/portfolio/*`, `/api/v1/treasury/*` - `/api/v1/market/*`, `/api/v1/swap/*`, `/api/v1/bridge/*`, `/api/v1/gas/*` - `/api/v1/risk/*`, `/api/v1/wallet/*`, `/api/v1/tax/*`, `/api/v1/foresight/*` - `/api/v1/agentic/*`, `/api/v1/watchers/*`, `/api/v1/monitors/*` - `/api/v1/economy/*` (x402, ACP, SKUs), `/api/v1/governance/*`, `/api/v1/defi/*` - `/api/v1/knowledge/*`, `/api/v1/safety/*`, `/api/v1/compliance/*` - `/api/v1/notifications/*`, `/api/v1/inbox/*`, `/api/v1/digests/*`, `/api/v1/webhooks/*` - `/api/v1/audit/*`, `/api/v1/observability/*`, `/api/v1/billing/*`, `/api/v1/entitlements/*` - `/api/v1/teams/*`, `/api/v1/privacy/*`, `/api/v1/storage/*`, `/api/v1/search/*` - `/api/v1/status/*`, `/api/v1/jobs/*`, `/api/v1/exports/*`, `/api/v1/tags/*`, `/api/v1/flags/*` - `/api/v1/insights/*`, `/api/v1/marketplace/*`, `/api/v1/sla/*`, `/api/v1/contracts/*`, `/api/v1/workflows/*` - Legacy convenience: `POST /api/command`, `POST /api/planner`, `POST /api/simulate`, `POST /api/swap/*`, `GET /api/vault/balances`, `POST /api/squads/propose`, `POST /api/research`, `POST /api/tools/crawl`, `POST /api/tools/browser`, `GET /api/yield/opportunities`, `POST /api/dreyv-gateway/*` Auth: `Authorization: Bearer `. OAuth2 + API key flow documented at `/developers`. ## 9. Discovery & agent-ready endpoints - `/.well-known/mcp.json`: MCP server discovery - `/api/mcp/manifest`: full MCP tool catalog (public) - `/.well-known/ai-plugin.json`: OpenAI plugin manifest - `/.well-known/acp.json`: Agentic Commerce Protocol descriptor - `/.well-known/ucp`: Universal Commerce Protocol profile - `/.well-known/api-catalog`: RFC 9727 catalog - `/.well-known/x402`: x402 payment capability (advertised) - `/openapi.json`: OpenAPI 3.1 spec - `/llms.txt`: short canonical context - `/llms-full.txt`: this file (full single-request context) - `/api/llms.txt`, `/developers/llms.txt`: scoped agent contexts - `/pricing.md`: machine-readable pricing - `/index.md`: markdown homepage fallback - `/developers`: developer portal - `/developers/mcp`: MCP setup + client snippets - `/developers/api`: REST reference - `/developers/agents`: agent integration matrix (Claude, Cursor, elizaOS, Continue, custom) - `/developers/docs/[openapi|ai-plugin|acp|ucp|api-catalog|x402|llms-txt|llms-full|pricing-md]`: one page per spec - `/compare`: competitive comparison ## 10. Brand & legal - Brand: **dreyv** (pronounced "DRAVE"). Domains: dreyv.com, dreyv.app, dreyv.finance. - Company: dreyv (independent): founded 2025, headquartered in Nicosia, Cyprus. - Primary deployment: https://dreyv.com (Cloudflare Containers). - Legal: `/legal/terms`, `/legal/privacy`, `/legal/cookies`, `/legal/refunds`. - Social: https://x.com/dreyvapp · https://www.linkedin.com/company/dreyv/ ## 11. Contact - Product: founders@dreyv.com - Security: security@dreyv.com - Press / partnerships: hello@dreyv.app