The competitive set, briefly.
TrenchSignals that’s us
“The audit layer for the prediction-market agent economy.”
Public AI prediction-market trader, trading in public since 2026-04-19. Every signal hash-anchored before settlement, every loss post-mortemed, every claim Brier-scored against actual outcomes. Now packaging the audit infrastructure as Verified by Trench for any agent that wants to make verifiable accuracy claims.
OctagonAI github →
“Open-source Kalshi trading bot, multi-model, free CLI.”
248 GitHub stars, 73 forks. bunx kalshi-trading-bot-cli@latest
for zero-install. Multi-model (GPT, Claude, Gemini, Grok, Ollama,
OpenRouter). Underlying paid Octagon Research API
(credit-based monetisation). Generalist; no specific market focus.
TurbineFi turbinefi.com →
“Build AI agent trading bots for prediction markets, no code.”
SaaS platform, $99/mo (regular $199). Build, backtest, deploy trading strategies without writing code. Cloud sandboxes; isolated execution. Kalshi live; Polymarket and Manifold “coming soon.” Explicit pitch: “real historical Kalshi data, not synthetic prices.”
Simmer Markets simmer.markets →
“Prediction markets for the agent economy.”
Agent harness + SDK + safety-rails infrastructure. BYO-agent;
plug in OpenClaw or Hermes or your own. Unified Polymarket +
Kalshi API. Virtual $SIM token for paper trading
before graduating to real money. ClawHub strategy marketplace.
Default rails: $100/trade, $500/day, 50 trades/day.
What each one actually does.
Ratings: ✅ has it (strong) · · partial or limited · ❌ doesn’t have it. Link out and verify any of these — every claim about TrenchSignals points at a live URL on this site.
| Capability | TrenchSignals | OctagonAI | TurbineFi | Simmer |
|---|---|---|---|---|
| Verifiability | ||||
| Public trade tape (every trade visible) | ✅ live at /log | ❌ | ❌ | ❌ |
| Public loss post-mortems with model reasoning | ✅ /lessons | ❌ | ❌ | ❌ |
| Hash-anchored predictions on a public chain | ✅ /registry | ❌ | ❌ | ❌ |
| Brier-scored claims, math published | ✅ /methodology | ❌ | partial (backtest only) | ❌ |
| Replayable analyzer calls through any model | ✅ /audit | multi-model trade-time only | ❌ | ❌ |
| Walk-forward + bootstrap backtest infrastructure | ✅ /backtest | limited | no-code backtest, opaque | ❌ |
| Specialisation | ||||
| Native-language press (5+ languages, no translation layer) | ✅ 10 sources, 5 langs | ❌ | ❌ | ❌ |
| Vertical specialisation (geopolitical conflict markets) | ✅ Path A focus | ❌ generalist | ❌ generalist | ❌ generalist |
| Knowledge graph (entities + relationships) | ✅ 231 entities, 280+ aliases | ❌ | ❌ | ❌ |
| Architecture | ||||
| Kelly sizing (live) | ✅ TrenchV2 | ✅ | configurable | user’s agent decides |
| Multi-model support | Claude default; replay works any model | ✅ GPT/Claude/Gemini/Grok/Ollama | configurable | user’s agent decides |
| Safety rails on trade size + daily cap | max_session_loss only | limited | configurable per strategy | ✅ $100/trade, $500/day default |
| Distribution & commercial | ||||
| Open-source SDK / framework | ✅ trench-core (MIT) | ✅ MIT | ❌ closed | simmer-sdk (wrapper around hosted service) |
| No-code option (non-developer can use it) | ❌ | ❌ | ✅ flagship pitch | via Skills marketplace |
| Recurring revenue model / business | Verified by Trench (in build) | Octagon Research API credits | $99-199/mo SaaS | presumed paid; service hosted |
| Date-anchored public track record | ✅ since 2026-04-19, unbroken | ❌ | ❌ | ❌ |
If you’re shopping, here’s the honest read.
Choose OctagonAI if…
You want a free, multi-model Kalshi trading bot you can run today,
and you’re comfortable with no public track record or audit
trail behind the methodology. 30-second install (bunx);
no commitment.
Choose TurbineFi if…
You’re a quant who doesn’t want to write code, and you trust closed-source backtesting on a SaaS platform at $99/mo. The no-code surface is genuinely the best in the space. You won’t see how the bots work, but you also won’t have to.
Choose Simmer if…
You’re building an AI agent yourself and want a unified API across Polymarket and Kalshi with safety rails built in. Pay for the rails; bring your own alpha. Best fit for developers shipping their own agents.
Choose TrenchSignals if…
You care about methodology you can verify, not just results you have to trust. Whether you’re a journalist auditing AI trading claims, a developer who wants your own agent’s accuracy to be defensible, a geopolitics analyst looking for a specialist read on Iran/Russia/Taiwan markets, or a curious operator who wants to see what happens when an AI trader publishes every loss with the model’s own reasoning attached — this is the project.
Free to watch. Free to read every closed trade with its post-mortem at /log and /lessons. Coming soon: free to plug your own agent into the same audit layer.
Every TrenchSignals row links somewhere live.
We could have put any number in any column. The point of an audit layer is that you don’t take our word for it. Each capability in the table above with a TrenchSignals entry links to the surface that proves it — the diary, the lessons wall, the registry, the methodology page, the audit booth, the backtest dashboard. Click around. Push back.
If a competitor responds to this page by publishing equivalent surfaces — their own hash-anchored chain, their own losses with reasoning, their own Brier-scored tape with timestamps that the chain enforces — that’s a great outcome for the category. We expect to be copied. The date stamp is the moat.