Roadmap

Every item on this page carries one of four statuses. Nothing is described as live unless a route, engine or deployed contract in the repository actually implements it.

For what each area means in the product, see Docs. For the reasoning behind it, see Whitepaper → long-term ecosystem vision.

The MPGR Agent on Home: research, reason and prepare onchain actions, with every write confirmed by the user.

• LIVEAgent workspace on Home — chat, status, prompt suggestions, live Base market tape
• LIVEIntent detection against a closed intent list with deterministic replies
• LIVESmart actions and follow-up prompts derived from live account context
• LIVEPrepare-only proposals: Base transfer, swap, B20 order, x402 payment
• LIVEDeterministic fallback engine when every network provider fails
• IN PROGRESSBroader tool coverage and richer multi-step research answers
• PLANNEDPersistent, user-controlled agent memory across sessions
• LONG-TERM VISIONLong-horizon personal agent that plans across sessions inside user-set limits

What the Agent is allowed to do on its own — today, nothing that moves value.

• LIVEEvery value-moving action requires explicit confirmation and a wallet signature
• LIVENo autonomous execution path; AgentKit signing and auto-payment actions denied server-side
• LIVEPrepare-only AgentKit configuration on Base mainnet with a read-action allowlist
• PLANNEDUser-configurable spend caps and per-action allowance limits
• PLANNEDPolicy profiles — read-only, prepare-only, bounded-execute
• LONG-TERM VISIONDelegated, revocable agent permissions with onchain-enforceable budgets

Read-only facts the Agent can retrieve and reason over — from Base, from MPGR contracts and from the app's own ledgers.

• LIVEWallet, token, portfolio and Base research tools
• LIVE$MPGR market endpoint, with explicit DATA_UNAVAILABLE instead of invented prices
• LIVEYield tools normalised from live MPGR staking on-chain data
• LIVECoinbase B20 tokenized-stock research and contract verification
• IN PROGRESSDeeper portfolio analytics and history surfaced through the Agent
• PLANNEDWatchlists, alerts and additional market data providers
• LONG-TERM VISIONResearch copilot with saved theses, scenarios and cited sources

Swaps and tokenized-stock orders on Base, always prepared and confirmed before anything is signed.

• LIVECoinbase CDP Trade API with a 0x Swap API fallback
• LIVEB20 tokenized-stock orders through Aerodrome Slipstream USDC pools
• LIVESlippage bounds, quote freshness, risk facts and receipt verification
• LIVEQuotes bound to the authenticated session wallet
• PLANNEDLimit orders, recurring buys and scheduled execution — all confirm-gated
• PLANNEDPrice alerts and execution analytics
• LONG-TERM VISIONStrategy layer where the Agent proposes rule-based execution you approve

Turning an answer into the next correct step, without ever letting a model choose a destination.

• LIVEContext-aware action cards, highlight chips and follow-up prompts
• LIVEClosed navigation whitelist — the model selects an intent, never a route string
• PLANNEDMulti-step action plans (research → prepare → confirm) in one flow
• LONG-TERM VISIONUser-authored automations with explicit budgets and an audit trail

A pluggable provider stack with a deterministic safety net, so no single model is a single point of failure.

• LIVEProvider interface with Gemini (default), NVIDIA NIM and OpenAI implemented
• LIVEDeterministic on-device fallback provider
• LIVETask-based routing, guardrails, timeouts, circuit breaker and diagnostics
• LIVEPrompt/output budgets and a daily AI token budget
• PLANNEDAnthropic and self-hosted providers behind the same interface
• PLANNEDPer-task provider selection with cost and quality telemetry
• LONG-TERM VISIONOpen provider marketplace where third-party models register behind the guardrail stack

Agents that can be paid for work — and eventually pay each other — inside limits a human set.

• LIVEUsers can pay x402-enabled resources through the Agent with explicit confirmation
• PLANNEDMetered agent usage and MPGR-addressable agent services
• LONG-TERM VISIONAgents that discover, quote and pay each other with user-set budgets and receipts

The payment rail for machine-to-machine commerce, scoped hard to Base and to explicit consent.

• LIVEx402 discovery, server-registered terms, confirm, EIP-3009 sign, bound submission
• LIVEBase mainnet only, exact scheme only, known-asset EIP-712 domain required
• LIVENo silent payment — AgentKit auto-payment actions are denied server-side
• IN PROGRESSWider resource compatibility and clearer failure messaging
• PLANNEDReceipt history and exportable payment records
• PLANNEDRecurring / subscription-style payments behind explicit caps
• LONG-TERM VISIONMPGR HUB as an agentic-payment gateway for Base services

A registry where models, tools and services can be listed, priced and paid for.

• LIVETool registry with schema, risk level, timeout and confirmation requirements
• PLANNEDListing schema and pricing for third-party models, tools and services
• PLANNEDReputation and ratings derived from verified usage
• LONG-TERM VISIONAI / service marketplace with x402-metered access, funded from the AI ecosystem treasury line

A registry-driven games hub. Titles ship only when their route actually exists.

• LIVEGames hub with a registry-driven catalog
• LIVEComing-soon titles clearly labelled, never presented as playable
• PLANNEDAdditional playable titles (Clicker, Memory Challenge, Space Shooter, 2048 Daily and others)
• PLANNEDShared progression: XP, season points and campaigns across games
• LONG-TERM VISIONMultiplayer modes, tournaments and cross-game seasons

The flagship title: server-issued sessions, authoritative replay verification, and financial payouts kept fail-closed.

• LIVEPlayable endless runner with server-issued sessions and heartbeats
• LIVEIn-process deterministic authoritative replay — seed plus input trace replayed tick-for-tick
• LIVEDrift-tolerant timing checks, rate and idempotency gates
• LIVE8 XP per completed run, capped at 10 XP-earning runs per day
• LIVEWeekly stats and leaderboard identity bound to the authenticated wallet
• IN PROGRESSAnti-cheat hardening beyond deterministic replay and heartbeat/timing gates
• IN PROGRESSEnablement review for financial game rewards (two independent operator gates, fail-closed)
• PLANNEDSeasonal game modes and cosmetic progression
• LONG-TERM VISIONCertified competitive play with real-time leaderboards

A Reward Hub over real on-chain claims, with the rest of the program surfaced as providers come online.

• LIVEReward Hub: summaries, categories, claim history, achievements, season preview
• LIVEOn-chain claiming through the deployed MPGRRewardVault (claim / claimMultiple)
• LIVEServer-owned XP ledger, daily check-in, streaks and levels
• LIVESeason points on UTC monthly seasons plus a Season Pass reward track
• PLANNEDAdditional reward providers behind the hub (AI, premium, airdrop)
• LONG-TERM VISIONUnified earn graph where every product action maps to a transparent reward rule

Single-sided MPGR staking against the deployed Base contract — no lock term, claim any time.

• LIVEStake, unstake, claim rewards and exit via MPGRStaking on Base
• LIVELive APR, total staked and pool state with short cache TTLs
• LIVEStaking history with progressive backward backfill
• IN PROGRESSBackfill performance work within provider-tier RPC limits
• PLANNEDAdditional pools and richer staking analytics
• LONG-TERM VISIONGovernance-directed emission schedules

Time-locked MPGR on the immutable V1 contract — the commitment primitive behind Premium and Holder Score.

• LIVEcreateLock / withdraw / earlyUnlock against MPGRTokenLock V1 on Base
• LIVEFixed 10% on-chain early-unlock penalty, computed and executed by the contract
• LIVELocked MPGR drives Premium tier and Holder Score
• PLANNEDLock analytics and lock-based campaigns
• LONG-TERM VISIONLock-weighted governance power

What $MPGR does today, and what governance could add later — without changing the fixed supply.

• LIVEStaking yield, lock / holder commitment, vault claims
• LIVEGame and season incentives when the operator gates are on
• LIVEReferral and quest budgets; the economic surface the Agent operates around
• PLANNEDCommunity governance — treasury reallocation within the 100M cap, emission rates, campaign budgets, emergency reserve
• LONG-TERM VISIONOnchain proposal lifecycle with voting weight from Holder Score

Progression surfaces that must stay honest — so ranking and attribution are server-owned.

• LIVEAuthenticated referrals with self-referral blocking and abuse logging
• LIVEGlobal leaderboard sourced from the server ranking, never a client score
• LIVECampaigns: config-driven events with their own points ledger and leaderboard
• IN PROGRESSReferral sybil resistance
• PLANNEDRicher quest engine with onchain and offchain verification
• LONG-TERM VISIONReputation graph with sybil-resistant identity

Base is the home chain. There is no multi-chain runtime today and none is promised.

• LIVEBase mainnet only (8453) across every tool, contract and route
• LIVECoinbase CDP Trade API, B20 via Aerodrome Slipstream, USDC on Base
• LIVEFarcaster Mini App connector and Basename resolution
• IN PROGRESSDeeper Base ecosystem campaigns and integrations
• PLANNEDMore Base-native protocols behind the same prepare / confirm boundary
• LONG-TERM VISIONBase-first onchain operating system; cross-chain remains out of the current runtime

What is already enforced in code, and the independently verified guarantees still missing.

• LIVESIWE sessions, HMAC session cookies, request guards, rate limits and daily AI budget
• LIVERead/write tool separation, risk levels and mandatory confirmation
• LIVESanitized errors — no raw provider, RPC, Redis or stack-trace output to clients
• LIVECI: lint, typecheck, unit and security tests, high-severity audit script, Foundry contract tests, production build
• LIVEFail-closed game settlement without both operator gates
• IN PROGRESSIndependent smart-contract audit
• IN PROGRESSVault-level settlement idempotency or a durable outbox
• IN PROGRESSPerformance and production-funding review
• PLANNEDPublic bug bounty and recurring audits
• LONG-TERM VISIONContinuous onchain monitoring with automated circuit breakers

Opening the modules that already exist, without weakening the boundaries that make them safe.

• LIVEInternal API surface for market, trade, transfer, x402, XP, referral, leaderboard, campaigns and games
• LIVEPublic GitHub repository with architecture, security and runbook documentation
• PLANNEDPublic API keys and documented endpoints for partners
• PLANNEDSDK and an embeddable agent surface
• LONG-TERM VISIONPartner ecosystem building on MPGR HUB modules

The runtime, the deployment model and the operational maturity work.

• LIVENext.js App Router on Vercel, GitHub-connected deploys
• LIVERedis / KV-backed sessions, XP ledger, referrals, leaderboard and game allocation
• IN PROGRESSGame art optimisation — responsive sizes, lazy loading, preloading
• PLANNEDMobile wrapper for the responsive web app
• PLANNEDRicher observability and operational runbooks
• LONG-TERM VISIONModular service platform with self-hostable components

Stating what MPGR HUB is not doing is part of keeping this page honest.

• Cross-chain execution. Base is the home chain and the only configured runtime. Cross-chain support is not committed.
• Custodial trading. The Agent never holds keys, never holds a balance for you, and never signs on your behalf.
• Paid subscriptions. Premium is a function of locked MPGR, not a recurring payment.
• New token supply. Maximum supply stays fixed at 1,000,000,000 MPGR. Rewards come from the community treasury.
• Retail tokenized-stock minting. Primary mint and redeem are Authorized Participant only; MPGR HUB implements no retail mint API.

Related: Docs → overview · About → product direction · $MPGR → tokenomics