# Canonopy Decisions > Canonopy Decisions: send us the JSON you send Jev. In minutes, get your own 1 MB model. For games and rule-based decisions, it beats Jev from day one. For text, it beats Jev clearly once it learns from your decisions. It is your own model for a decision you make over and over (route a ticket, approve a refund, flag a transaction, pick an NPC's move), asked in the same request and response format as TypeSafe's Jev (System One): Choice, Score and Noul questions. Send one real example of your state and your questions, exactly as you send them to Jev, to POST /v1/build (plus 5-20 more real states in `examples` and your rules in plain words if you like: no past decisions or answers needed; a few examples of what you send Jev make it better). Your model is ready in minutes, with nothing to do in between; review how it decides on about 20 examples any time (optional). Then switch the base URL and model: POST /v1/decide (alias /v1/systemone) keeps taking the same JSON. Your rules are enforced on every decision; the model is under 1 MB (text models share one 34 MB reader) and runs on our servers or yours, even offline. Built automatically with zero data it beat Jev in Doom (45 kills to 39, 2 deaths to 5, at the same decision rate) and Snake (42.0 to 40.0 food, never died). On text it is close to Jev on day one for short messages (bank: 84.6-86.8% vs 86.6%) and pulls clearly ahead as it learns from your decisions: 91.7% after about 400 reviewed cases, about 95% with the bank's history; on consumer complaints with history, 77.7% vs 65.0%. Past cases become the main examples, rules found in your history are proposed (optional to confirm), and every retrain (about a minute) is your call. Each decision model is called a domain in the API (/v1/domains). A 5-day free trial starts at your first model or first decision, no card; then $20 a month per workspace, flat, with unlimited decision models, decisions and retraining and up to 30 builds a month (5 a day, 3 during the trial). Your models are yours to keep. API base URL: https://api.canonopylabs.com. Auth: `Authorization: Bearer ` (keys at https://canonopylabs.com/console/keys). The API description at https://api.canonopylabs.com/openapi.json explains every endpoint and field, with a "start here" workflow. Coding agents can also use the hosted MCP connector at https://api.canonopylabs.com/mcp (Streamable HTTP, the same key in an Authorization header, nothing to install), or the local `canonopy mcp` server: see https://canonopylabs.com/docs/coding-agents.md. No package is needed to use the API; `canonopy-runtime` runs a downloaded model offline. The agent skill (https://canonopylabs.com/skill/SKILL.md) says when and how to use Canonopy Decisions, step by step. ## Get started - [Introduction](https://canonopylabs.com/docs.md): Send us the JSON you send Jev. In minutes, get your own 1 MB model. For games and rule-based decisions, it beats Jev from day one. For text, it beats Jev clearly once it learns from your decisions. - [Quick start](https://canonopylabs.com/docs/quick-start.md): Send the JSON you send Jev to POST /v1/build and switch over: your own model in minutes, no data needed. Review how it decides any time (optional). Curl, Python, the CLI or the console. - [Start with no data](https://canonopylabs.com/docs/start-with-no-data.md): Send the same JSON you send Jev to POST /v1/build and get your own model back in minutes: no past decisions or answers needed. Paste a Jev request as is, add 5-20 more real states and rules in plain words, then keep sending the same JSON. Review how it decides any time (optional). - [Build with your coding agent](https://canonopylabs.com/docs/build-with-your-coding-agent.md): Let your coding agent (Claude Code, Cursor, Codex) build your model through our MCP connector: it brings the JSON you send Jev, answers a few questions about it from your code and logs, and your model is ready in minutes. - [Answer a few questions](https://canonopylabs.com/docs/answer-a-few-questions.md): Every build asks a few questions about the JSON you sent: what fields mean, units and ranges, where two choices meet, real examples, unwritten rules. Answers are optional and make a better model. - [What you need to bring](https://canonopylabs.com/docs/what-you-need-to-bring.md): Nothing but your Jev request: one real example of your state and your questions. Your decisions make it better: past cases, answers to unsure cases, recorded play. ## Concepts - [Questions](https://canonopylabs.com/docs/questions.md): Choice, Score and Noul questions in Jev's format, and what each answer holds. - [Decisions and rules](https://canonopylabs.com/docs/decisions-and-rules.md): The decision block: the final action with your rules applied, which rules blocked what, and act / review / ask. - [Confidence](https://canonopylabs.com/docs/confidence.md): Calibrated confidence, the automatic safety bar, and how to use confidence in your code. - [Languages](https://canonopylabs.com/docs/languages.md): Reading about 100 languages: strong languages, double-checked messages and the language field. - [Day one](https://canonopylabs.com/docs/day-one.md): Answers from your first call: your day-one backend (your Jev key, an OpenAI-compatible model or ours) answers a model or question you haven't built yet, until it has a model of its own. ## Train and improve - [Make it better with your history](https://canonopylabs.com/docs/with-your-history.md): The same builder with your past cases: send them in POST /v1/build or upload them, and your cases become the main examples, rules found in them are proposed for you to confirm, contradicting cases wait for your review, and your model is measured on your own held-back cases. - [Rules found in your history](https://canonopylabs.com/docs/rules-found.md): Rules found in your past cases, each with its support, agreement and what confirming it changes: confirm, reject or make it hard. And past cases that contradict your rules, flagged for you to review: follow the rule, keep, or drop. - [Improve your model](https://canonopylabs.com/docs/improve-your-model.md): Advice on what would help most, a targeted retrain with a before and after, rule changes that retrain straight away and say what they change, and marking a decision wrong. Every retrain is your call. - [Playtest your model](https://canonopylabs.com/docs/playtest-your-model.md): canonopy playtest: your own game plays episodes with your finished model on your machine, and only the numbers it prints come back: scores, deaths, stuck moments. - [Shadow mode](https://canonopylabs.com/docs/shadow-mode.md): Run your model beside Jev or your current process on real requests without acting on it; review the few places they disagree, and spot-check a few confident decisions each week. - [Training and reports](https://canonopylabs.com/docs/training-and-reports.md): Set up a decision model by describing it, upload past cases, train a version in about a minute, and read its report. Or build one from the JSON you send Jev. - [Outcomes and options](https://canonopylabs.com/docs/improving.md): Report outcomes, work the unsure queue, add options and clarify boundaries: each is one call. - [Running it yourself](https://canonopylabs.com/docs/running-it-yourself.md): Download a version and run it on your own servers or offline with canonopy-runtime. ## Build - [Use with coding agents](https://canonopylabs.com/docs/coding-agents.md): Connect Claude Code, Cursor, Codex and other coding agents to Canonopy Decisions with the hosted MCP connector (nothing to install), the local MCP server, the agent skill, or llms.txt and Markdown docs. - [Use your model in Godot](https://canonopylabs.com/docs/godot.md): The Canonopy addon for Godot 4: your downloaded model decides inside your game, from GDScript, offline, with the same decisions as the hosted API. - [Using with LLMs](https://canonopylabs.com/docs/using-with-llms.md): Put the rule-bound decision in your own model and let your LLM do the talking: support bots, guardrails, agents. - [Patterns](https://canonopylabs.com/docs/patterns.md): Jev's four patterns (fan-out, confidence routing, weighted scores, handlers), unchanged on our answers. ## Cookbooks - [Doom, zero data](https://canonopylabs.com/docs/cookbooks/doom-zero-data.md): A Doom player built automatically from the game's Jev request, with no recorded play, in about 4 minutes: at the same decision rate as Jev, 45 kills to 39 and 2 deaths to 5 over 13 games, and 12 wins of 13 against Jev's aggressive style. - [Snake, zero data](https://canonopylabs.com/docs/cookbooks/snake-zero-data.md): The Snake game from Jev's own demo, played by a model built automatically from the demo's Jev request, with no recorded moves, in about 3 minutes: 42.0 food to Jev's 40.0 at equal steps, never died. - [Bank support, day one](https://canonopylabs.com/docs/cookbooks/bank-zero-data.md): A bank's support desk built from its Jev request in minutes, then improved with reviewed cases and history: close to Jev on day one (85.5% against 86.6% on 3,080 real messages), 91.7% after about 400 reviewed cases, about 95% with the bank's history. - [Bank support, with history](https://canonopylabs.com/docs/cookbooks/bank-support.md): 3,080 real bank messages, head to head with Jev: 96.0% against 86.6% with the bank's history, and how to build a model like it from your past cases. - [Doom, recorded play](https://canonopylabs.com/docs/cookbooks/doom.md): A Doom player built through the API from recorded play: 56 kills and 0 deaths against Jev's 34 and 6, over the same 13 live games. - [Snake, recorded moves](https://canonopylabs.com/docs/cookbooks/snake.md): The Snake game from Jev's own demo: a decision model over the game state, learned from recorded moves, that picks each move, called like Jev or run inside the game loop. - [Ticket triage, day one](https://canonopylabs.com/docs/cookbooks/ticket-triage.md): A helpdesk with no history: start in day-one mode and let each question graduate. ## Reference - [API reference](https://canonopylabs.com/docs/api-reference.md): Every endpoint: decide, build from a Jev request, rules found in your history, improving a model, decision models and set-up, past cases, training, versions, outcomes, day-one and account. - [CLI reference](https://canonopylabs.com/docs/cli-reference.md): The canonopy command line, its local MCP server for coding agents, and the runtime CLI. All optional: the API is plain HTTP. - [Errors and limits](https://canonopylabs.com/docs/errors-and-limits.md): Error shapes and types, the fair-use rate limit, build limits, and size limits. ## Skill - [Canonopy Decisions skill](https://canonopylabs.com/skill/SKILL.md): an agent skill (SKILL.md) for Claude Code and other coding agents: when a decision model fits, every step with its MCP tool and API call, reading answers, switching from Jev, pricing and errors - [The skill as a zip](https://canonopylabs.com/dl/canonopy-decisions-skill.zip): the canonopy-decisions/ folder with its reference and example files, to unzip into a skills folder (or run `canonopy skill install`) ## API - [OpenAPI description](https://api.canonopylabs.com/openapi.json): every operation (what it's for, when to use it, what comes before and after, its errors) and every field, with examples - [Interactive API docs](https://api.canonopylabs.com/docs): the same description, browsable ## Optional - [Every docs page in one file](https://canonopylabs.com/llms-full.txt): all of the pages above, concatenated as Markdown