# Conclave

Plain-language brief for investors and language models. Figures below are from public HTTP endpoints or from this repository’s source. They are not audited financial statements. Nothing here is an offer, a profit forecast, or financial advice.

**As of 2026-11-12T01:21:55Z** (UTC). Live fields were read from `https://api.conclavedao.fun` around that time. `GET /api/performance?era=all` reported `generatedAt` `2026-11-12T01:21:55.502Z`.

---

## One-liner

Conclave is a live Solana trading loop: four software agents nominate and score which token the treasury buys, the engine executes on Jupiter, and realized profit is split by a published treasury-growth rule. The project token is `$CONCLAVE`.

---

## Problem

On-chain “treasury trading” products are usually opaque (who picked the trade, what the exit was, whether holders were paid) or fully manual (holders vote slowly while memecoins move in minutes). Conclave’s stated product is a clocked round with a public ledger of fills, PnL, and payouts, and an agent council instead of a live holder ballot.

Whether that design earns money is an empirical question. Cumulative realized trading PnL on the public books is negative (see Live traction).

---

## Product & how it works

### Round cycle

States: `NOMINATING` → `VOTING` → `HOLDING` → `DISTRIBUTING` → next round (README; `web/components/Docs.tsx`).

Live phase lengths from `GET /api/config`: `nominationHours` 0.25, `votingHours` 0.25 (15 minutes each). `agentCouncilEnabled` true, `holderVotingEnabled` false, `paperMode` false, `agentLearningEnabled` true.

Holder nominate/vote HTTP routes stay off unless `HOLDER_VOTING_ENABLED=true`. That flag turns the council off and restores a legacy holder ballot (README). It is not the live path.

### AI council

Four equal personas scan trending Solana mints that already pass safety checks (minimum liquidity, 24h volume, pair age, Jupiter routability; not native SOL and not `$CONCLAVE`):

| Agent | Id | Role |
|---|---|---|
| Vector | `momentum` | Momentum |
| Depth | `liquidity` | Liquidity / exitability |
| Sentinel | `risk` | Risk-averse |
| Fade | `contrarian` | Contrarian / pullback |

Each agent nominates one mint and, in `VOTING`, scores every validated nomination 0–100. Weight equals that score. 0 is an abstention. Ties break by supporter count, then 24h volume, then mint address. Top picks with a positive tally form the buy list (50/30/20, renormalized if fewer survive). An empty tally can fall back to an edge-score ranker only if `ALLOW_UNVOTED_FALLBACK` is on (README).

A council pick can still be dropped at buy time (momentum, volume, spend cap, price-impact, sellability). Live public knobs from `GET /api/config` include `maxBuySlots` 1 and `antiChaseEnabled` true. Exact liquidity floors and spend caps are not all on the public config (docs say several numbers are unpublished).

After the buy, names and scores are published. Before the buy, `GET /api/rounds/active` returns `disclosure.ballot: "withheld"` (observed 2026-11-12).

Learning is on in production (`GET /api/learning` `enabled` true, `steering` true). Closed live trades update each agent’s vote weight. At the snapshot, Fade was `benched: true` with `voteWeight` 0; the other three were not benched (`GET /api/learning`, `GET /api/agents`).

### Trading

Buys and sells go through Jupiter from a custodial treasury hot wallet. `PAPER_MODE` is false in production (`GET /health`, `GET /api/config`). The hold monitor re-quotes open legs on a short interval (code default 15s in `api/src/lib/riskMath.ts`). A missing/stale quote can force an exit. Failed sells retry on a slippage ladder.

### Exits (including stop, profit-lock, trail, dead zone)

New positions lock take-profit, stop, and max hold from the council before the buy. An open round keeps the plan it was bought with. Those distances are **not** on the public API: `GET /api/rounds/active` `disclosure.exit` is `"hidden"`; public `GET /api/config` omits the `exit` object (see `PUBLIC_CONFIG_KEYS` in `api/src/lib/publicRedact.ts`).

From source, not from a live public field:

- If a round has no stored plan, the code fallback is +3% take-profit and −3% stop (`UNSTORED_TAKE_PROFIT_BPS` / `UNSTORED_STOP_LOSS_BPS` = 300) and `MAX_HOLD_HOURS` default 24h (`api/src/lib/agents/exitPlan.ts`, README).
- **Stop-loss:** a leg at or through its locked stop is sold; the basket can also exit on that stop. A buffer and extra slippage exist so the fill can land near the stop (`api/src/lib/riskMath.ts`).
- **Take-profit:** the basket can exit when combined PnL reaches the locked target.
- **Trailing stop:** per-leg trail exists in the trader. New-leg default arm is two-thirds of a 3% target (200 bps) with 40% giveback (`NEW_LEG_TRAIL_ARM_BPS`, `NEW_LEG_TRAIL_GIVEBACK_PCT`). Whether the live process has trail enabled is not on public `/api/config`.
- **Profit-lock:** once peak PnL reaches an arm (code default +1.2%), a dip through a floor (code default −0.3%) can sell before the hard stop. Selling is off unless `PROFIT_LOCK_ENABLED` is exactly `true` or `1`; shadow logs still run (`api/src/lib/riskMath.ts`). That env flag is not published.
- **Dead-zone lock (live as of 2026-11-12):** code constant `DEAD_ZONE_LOCK_ENABLED = true` in `api/src/lib/riskMath.ts` (not env-driven). After a peak of **+120 bps**, if PnL falls to **≤ +40 bps** while peak is still below the trail arm (+200 bps default), the engine **sells** that leg with reason `dead_zone_lock`. Monitor precedence: stop-loss → profit_lock → trail → **dead_zone_lock** → basket TP/max-hold → no_progress (`planPositionExits`). Peaks that reach the trail arm defer to trail/SL. A separate replay grid in `deadZoneGiveback.ts` still records would-fires for research; that grid is not the live seller. Public `/api/config` still omits the `exit` object, so this description is from source, not from a public JSON field.
- **No-progress:** code default exits a stuck flat/red leg after 90 minutes if peak never reached the trail arm (`NO_PROGRESS_HOLD_MS`).

Closed rounds can show an outcome label (take profit, stop, trail, dead-zone lock (+0.4% floor), time). Open-round exit numbers stay hidden. Public receipts map stored `dead_zone_lock` to its own coarse reason instead of trailing.

### Risk controls (buy side)

From README and `api/src/lib/riskMath.ts` (code defaults; several are unpublished on `/api/config`): Jupiter routability, pool liquidity floor, 24h volume floor, pair age, max entry price impact (code 1%), round-trip loss estimate, spend cap as a fraction of treasury, anti-chase on extended 24h runs / 5m spikes, slot gate (`maxBuySlots` 1 on live config), gas reserve, exclude SOL and `$CONCLAVE`. These reduce some failure modes. They do not prevent losing trades.

---

## Live traction & performance

Sources: `GET /api/performance?era=all`, `GET /api/revenue`, `GET /api/rounds?receipts=1`, `GET /api/stats`, `GET /api/rounds/active`, `GET /api/token`.

**Closed trades with recorded PnL: 98** (`performance.summary.rounds`, `revenue.totals.roundsCompleted`, `GET /api/rounds?receipts=1&outcome=all` `total`). **Wins 38, losses 60.** Receipts: `outcome=wins` total 38, `outcome=losses` total 60. No flat closed trades in that split.

**Win rate:** `performance.summary.winRateBps` **3878** (38.78%). `revenue.totals.winRateBps` **3877** (integer floor of 38/98). Both are from the same 38/98 count.

**Cumulative realized trading PnL: −566,935,645 lamports = −0.566935645 SOL** (`performance.summary.totalPnlLamports`, `revenue.totals.tradingPnlLamports`). This number **includes losses**. It is not mark-to-market of `$CONCLAVE`. It is not treasury balance.

Other `GET /api/performance?era=all` `summary` fields:

| Field | Value |
|---|---|
| Total deployed | 22,782,346,330 lamports (22.782346330 SOL) |
| Avg win | +851 bps |
| Avg loss | −735 bps |
| Profit factor | 0.650817 (< 1) |
| Max drawdown | 710,077,975 lamports (0.710077975 SOL); 665 bps |
| Best closed round | #19 `INUINK`, +357,212,626 lamports, +12,541 bps |
| Worst closed round | #7 `familiars`, −239,701,474 lamports, −1,625 bps |

First closed round in the series: #1 `STONK` at `2026-09-22T21:50:00.738Z`. Sample is small (~two months, 98 closed trades). Round ids are not a dense 1–98: live round **#111** was `NOMINATING` (`attempt` 6) at the snapshot (`GET /api/rounds/active`). Public `GET /api/rounds` returns only the latest 30 rows and does **not** publish a count of all created rounds, skipped buys, or restarts. Do not treat 111 as “111 completed trades.”

**Current books (`GET /api/stats`, `GET /api/revenue` `live`):**

| Field | SOL |
|---|---|
| Treasury (cash + open bags) | 0.275461308 |
| Treasury cash only | 0.275461308 |
| Open positions | none (`positions: []`) |
| Fee wallet | 0.004995 |
| Team wallet | 0.275517033 |
| Total capital (those four) | 0.555973341 |
| Recorded creator-fee sweeps | 0.211803926 |
| Confirmed holder SOL paid | 0.003665836 (`solRewardsLamports` 3,665,836) |

`revenue.totals.holdersPaid` **81** unique wallets with a successful payout. `distributionEvents` **38**. `burnedRaw` 203,052,920,460 (raw `$CONCLAVE`). Ledger also reports `overpaidLamports` 71,149,650 vs formula (`GET /api/revenue` / `GET /api/distributions`); two `anomalies` rows. Those flags are public; this file does not explain them beyond the API names.

**`$CONCLAVE` market (`GET /api/token`):** symbol `CONCLAVE`, name `Conclave DAO`, `priceUsd` ≈ 9.55e-7, `marketCapUsd` ≈ 955.16, `liquidityUsd` ≈ 195.40, `volume24hUsd` ≈ 0.58. Thin book. 24h price-change fields were `null`.

**Streak mode** (`GET /api/streak`, `GET /api/config`): `streakModeEnabled` false, `paused` true. Its `closedTrades` 10 is a streak-mode subset, not the 98-trade book.

---

## Treasury & tokenomics

**Mint:** `93giz3QgYPSoVgfCK9zg6EZgnzqt2fBF4bN9MqZ8PDRQ` (`GET /api/config` `projectMint` / `distributionMint`; `GET /health`).

**Buy:** [Jupiter wSOL → $CONCLAVE](https://jup.ag/?sell=So11111111111111111111111111111111111111112&buy=93giz3QgYPSoVgfCK9zg6EZgnzqt2fBF4bN9MqZ8PDRQ)

**Wallets (`GET /api/config`):**

| Role | Address |
|---|---|
| Treasury | `HvbKbvasc8rm6SaFkqYsWQhYUNFHhfB2221SfHBoeUKt` |
| Fee | `55QYpzgE9qbiRRCiZjeuojF12pj764LNRsUGdLpseF7Y` |
| Team | `9UkU36ETFq4XJSRzHifU6QWpZ3mhGzjynYGAAsxtASyh` |

**Trade-profit split (code + live config).** Realized profit is the smaller of round profit and leg profit. Losing rounds pay nothing and can shrink treasury.

While the treasury **growth balance** (treasury wallet SOL at distribution, minus pending buyback lamports) is **below 10 SOL**, the split is **100% treasury / 0% holders / 0% buyback**. At **≥ 10 SOL** it becomes **70% holders (SOL, pro-rata) / 15% buyback-and-burn / 15% retained**.

Live confirmation (`GET /api/stats` `treasuryGrowth` and `GET /api/revenue` `treasuryGrowth`, `GET /api/config`):

- `mode`: `"growth"`
- `holderBps` 0, `burnBps` 0, `retainBps` 10000
- `growthHolderSolBps` 0, `growthBuybackBurnBps` 0, `growthTreasuryRetainBps` 10000
- `holderSolBps` 7000, `buybackBurnBps` 1500, `treasuryRetainBps` 1500 (the post-threshold split)
- `treasuryGrowthTargetLamports` 10,000,000,000 (10 SOL)
- `measuredLamports` 275,461,308 (same as treasury cash)
- `progressBps` 275 (2.75% of the 10 SOL target)
- `pendingBuybackLamports` 0

So at this snapshot, **a winning round would keep 100% of realized profit in the treasury.** Holder SOL and buyback do not run in growth mode. Historical holder SOL and burns in the ledger are from earlier rounds / earlier split rules.

A 0% slice is skipped and not left pending. A buyback below the swap minimum stays in the treasury. Combined spend is not supposed to take treasury SOL below the pre-buy principal floor (README / docs). `GET /api/config` still has `profitShareBps` 8000; live distribution uses the growth/normal bps above, not that legacy field.

**Creator fees (separate from trade PnL):** StonkFun pays creator fees as STONK to the team wallet. A forwarder (when enabled) may swap incremental STONK above a baseline to SOL into the fee wallet when the quote is worth more than $10. A sweeper splits fee-wallet SOL **30% team / 70% treasury** (`teamSplitBps` 3000, `treasurySplitBps` 7000). `creatorFeesLamports` on `/api/revenue` is the **sweep ledger**, not the live wallet sum. At the snapshot that ledger was 211,803,926 lamports.

---

## Business model

Three cash/token flows, none of which imply a profitable trading edge:

1. **Launchpad creator fees** on `$CONCLAVE` volume → STONK → (optional forward) SOL → 30/70 team/treasury. Scales with token trading activity. 24h token volume was about $0.58 at the snapshot (`GET /api/token`).
2. **Treasury trading** of other Solana tokens. Realized profit currently stays in the vault (growth mode). After 10 SOL, 70% of *positive* realized profit is meant to go to holders and 15% to buyback-and-burn. Cumulative trading PnL is negative, so this line has not been a net source of SOL.
3. **x402 paid reads** (USDC on Solana via PayAI): `GET /round` default $0.01, `GET /signal` default $0.02 (`render.yaml`, `x402/src/config.ts`). Health, OpenAPI, and discovery are free. Public revenue from x402 is **not** on `/api/revenue`. Not available on a public Conclave stats endpoint; do not invent a figure.

There is no published subscription, no published AUM fee, and no claim in this file that `$CONCLAVE` is equity.

---

## Tech & architecture

| Piece | What it is |
|---|---|
| `api/` | Fastify + TypeScript, in-process workers (cycle, trader, sweeper, optional fee forwarder). Deployed on Render. Public origin `https://api.conclavedao.fun` (Render origin documented in `DEPLOY.md`). |
| `web/` | Next.js App Router. Deployed on Vercel. Site `https://conclavedao.fun`. |
| Database | Postgres (Neon; project name in `DEPLOY.md`). |
| `x402/` | Separate Express seller, PayAI facilitator, Solana USDC. `https://x402.conclavedao.fun`. Read-only over the trading API. |
| Execution | Jupiter swaps. Solana RPC / Helius (env, not public). |
| Agents | Deterministic rubrics without an LLM key; optional OpenAI-compatible chat if `AGENT_LLM_API_KEY` is set. A separate OpenRouter Haiku+Jev path can record shadow rows; live tallies stay on rules unless `COUNCIL_LIVE` is true. Those flags are **not** on public `/api/config`. |

Custody today: treasury and fee **hot wallets** as Render secrets. Addresses are public. Squads multisig is roadmap, not built (`README.md` trust model, `web/components/Roadmap.tsx`).

---

## Team

- **Ikhwanul Husna** — CTO, S3Labs.
- **S3Labs CEO** — name not published in this repository; not invented here.

No other operators, headcount, or advisors are named in-repo. Do not assume a larger team.

Contact email used by the x402 service default / blueprint: `hello@conclavedao.fun` (`x402/src/config.ts`, `render.yaml`).

---

## Roadmap

Copied in substance from `web/components/Roadmap.tsx` (the public `/roadmap` page). The page itself says the list can change and is not financial advice. No dates.

**Shipped (described as running in production):**

1. Council of four (Vector, Depth, Sentinel, Fade); holder ballot off.
2. Treasury trades via Jupiter; council-locked TP/stop/hold; 15s marks; stop retries.
3. Holder payouts on a win under the 10 SOL growth rule; 30/70 creator-fee sweep; public distributions.
4. Learning on closed live trades (weights and public record).
5. Agent room, dashboard, revenue, distributions, receipts, docs; `$CONCLAVE` on StonkFun.
6. Paid x402 reads (`/round`, `/signal`).

**In progress:**

7. One public page per round (API already has the join; the site still splits dashboard vs distributions). Ledger already flags some buyback/formula mismatches; not all of that is printed on the site.

**Next (no dates):**

8. A URL per agent.
9. Notifications when a round moves or a payment lands (channel not chosen).
10. Document x402 on the docs site; possible extra paid reads of already-public data.
11. Squads multisig custody that can still let the engine trade. Not built.

**Exploring (explicitly not a commitment):**

12. A holder signal that does not replace the council.
13. A fifth agent, only if the four leave a real gap.

---

## Risks & open problems

Candid, not exhaustive.

1. **Negative cumulative realized PnL.** −0.566935645 SOL over 98 closed trades; profit factor 0.65; 60 losses vs 38 wins. The product has not demonstrated a durable edge on the public tape.
2. **Small sample and changing rules.** ~98 closed trades, round ids up to 111, multiple restarts, learning, gates, and split rules have changed. Past buckets are not one stationary process.
3. **Tiny treasury vs the 10 SOL switch.** ~0.28 SOL in the trading wallet vs a 10 SOL growth target (`progressBps` 275). Holder SOL and buybacks are off until that target. Getting there depends on deposits, creator fees, and/or future profitable trades — none of which are assured. Fees need `$CONCLAVE` volume; 24h volume was under $1 at the snapshot.
4. **Memecoin / execution risk.** Targets are often thin Solana tokens. Slippage, failed swaps, RPC faults, and gaps through stops have already shown up in the project’s own comments (e.g. price-impact vs stop in `riskMath.ts`). Stops and the dead-zone lock limit damage; they do not cap fills at the trigger price. The dead-zone lock became a live sell on 2026-11-12 (`DEAD_ZONE_LOCK_ENABLED`); the 98-trade tape above mostly predates that path.
5. **Concentration.** A round deploys a capped share of spendable treasury into up to `maxBuySlots` names (live config: 1). A bad round hits most of the vault.
6. **Custody / key risk.** Hot wallets on the API host. Compromise, misconfig, or a bad deploy can move funds. No live multisig. Holders do not control the treasury.
7. **Smart-contract / venue risk.** Jupiter, StonkFun/LaunchLab, token programs, and RPC providers can fail or change. `$CONCLAVE` itself is a memecoin-style mint with ~$955 reported market cap and ~$195 liquidity — high chance of large price moves or illiquid exits for holders.
8. **Token and regulatory.** This is a public crypto token plus an automated trading bot. Securities, commodities, money-transmitter, and tax treatment are jurisdiction-specific and unsettled. This file is not a legal opinion. There is no published registration. Calling the loop a “DAO” does not by itself create holder control (the council votes; holder voting is off).
9. **Conflicts.** Team wallet receives 30% of swept creator fees. Team and treasury keys are operator-controlled. Learning can bench an agent (Fade was benched at the snapshot).
10. **Opacity by design.** Exit distances, some gates, and model flags are hidden from the public API. That is documented; it also means an outside reader cannot fully reconstruct why a given fill happened from public JSON alone.
11. **Ledger anomalies.** Public distributions report `overpaidLamports` and `anomalies`. Treat the books as inspectable, not as proven-clean.
12. **x402 and fee forwarder** may be idle or thinly used; no public GMV figure was found.

---

## What we're asking for

TBD by founders.

---

## Links

- Site: https://conclavedao.fun
- This file: https://conclavedao.fun/company.md
- llms.txt: https://conclavedao.fun/llms.txt
- Docs: https://conclavedao.fun/docs
- Dashboard: https://conclavedao.fun/dashboard
- Track record: https://conclavedao.fun/performance
- Roadmap: https://conclavedao.fun/roadmap
- API: https://api.conclavedao.fun (`/health`, `/api/config`, `/api/stats`, `/api/revenue`, `/api/performance`, `/api/rounds`)
- x402: https://x402.conclavedao.fun
- X: public handle/URL in `web/lib/social.ts` and the site footer
- Token buy: https://jup.ag/?sell=So11111111111111111111111111111111111111112&buy=93giz3QgYPSoVgfCK9zg6EZgnzqt2fBF4bN9MqZ8PDRQ
- Mint: `93giz3QgYPSoVgfCK9zg6EZgnzqt2fBF4bN9MqZ8PDRQ`

GitHub is listed as private in `DEPLOY.md`. Do not treat it as a public source.

---

## Number sources (checklist)

| Claim | Source |
|---|---|
| As-of / performance generatedAt | `GET /api/performance?era=all` `generatedAt` |
| 98 closed trades, 38 wins, 60 losses, PnL, drawdown, PF | `GET /api/performance?era=all` `summary`; wins/losses also `GET /api/rounds?receipts=1` |
| 3878 vs 3877 win-rate bps | performance vs `GET /api/revenue` `totals.winRateBps` |
| Treasury / capital / growth split / paperMode | `GET /api/stats`, `GET /api/revenue`, `GET /api/config` |
| Mint, wallets, 15m phases, learning/council flags | `GET /api/config` |
| Live round 111 nominating | `GET /api/rounds/active` |
| Token price / MC / liquidity / volume | `GET /api/token` |
| Holder SOL paid, burns, holdersPaid | `GET /api/revenue` `totals` / `GET /api/distributions` |
| Fade benched | `GET /api/learning` |
| Buy URL / mint constant | `web/lib/buyUrl.ts` (matches config mint) |
| Exit fallbacks, profit-lock flag, live dead-zone lock | `api/src/lib/agents/exitPlan.ts`, `riskMath.ts` (`DEAD_ZONE_LOCK_ENABLED`, `DEAD_ZONE_LOCK_ARM_BPS` 120, `DEAD_ZONE_LOCK_FLOOR_BPS` 40, `planPositionExits`) |
| x402 prices | `render.yaml`, `x402/src/config.ts` |
| Team names | this brief’s instruction + in-repo mention of Ikhwanul; CEO name not in repo |

If a field is missing above, it was not on a public endpoint at the snapshot.
