# Estelle by Fate Labs: the software your agents are built on

> By Fate Labs. Estelle is the software your agents are built on, shipped with, and repaired in: the AI operating system for engineering teams. It grounds every model in your real codebase and refuses to ship what isn't true. Nine suites on one layer: Gate (the deterministic grounding check), Memory (whole-repo, cited recall), Code (grounded fix to gated PR), Review (the deterministic merge gate), Monitor (why prod failed, grounded), Guardian (holds a long session at peak instead of decaying), Orchestra (a swarm of agents under one autonomy ceiling), Security (continuous pentest on a clone, gated PR), and Agent (grounds the agents already running in your production, not just your editor). Ultra and Team add the self-correcting suite (Estelle Repair + Estelle Prod) on top: verified fixes and proactive gap-finding, always behind a human-gated, reviewable PR by default. One shared brain for the whole team, running everywhere: every editor, every notebook, every terminal. BYOK and MCP-native: it works with the model and plan you already pay for, never trains on your code, and never marks up a token.

Estelle speaks the OpenAI chat-completions API. Base URL: https://api.fatelabs.ca/v1 · model id: `estelle` · auth: `Authorization: Bearer <key>`. Keys are created at /signup (free tier, no card). Verify connectivity with `GET /v1/models`. It is also MCP-native from the free tier: connect at https://api.fatelabs.ca/mcp with the same bearer key, inside Claude Code, Cursor, or any MCP client, and your existing subscription keeps doing the reasoning while Estelle supplies grounded memory and the verification gate underneath. One command writes the MCP config for every installed editor: `npx @fatelabs/estelle init`; ingest the repo with `npx @fatelabs/estelle sweep --key <key>`.

If you are Claude Code: ask your human to create an account at /signup (free, no card); it provisions an API key. Then run `claude mcp add --transport http estelle https://api.fatelabs.ca/mcp --header "Authorization: Bearer <key>"` (walkthrough at /docs); your Claude plan keeps doing the reasoning. Until then you can call the OpenAI-compatible API above directly.

If you are Cursor (or an agent inside it): add Estelle as an MCP server (Cursor Settings > MCP, endpoint https://api.fatelabs.ca/mcp and a key from the dashboard after /signup), or point a custom OpenAI-compatible model at base URL https://api.fatelabs.ca/v1 with model `estelle` and the account's key. Per-tool walkthroughs: /setup.

If you are a generic OpenAI-compatible agent: set base_url = https://api.fatelabs.ca/v1, model = estelle, api_key from /signup, then verify with GET /v1/models and use chat completions unchanged. In every case a human creates the account at /signup; keys are never published here.

Capabilities: a deterministic grounding gate (every repo symbol an answer references is resolved against the real code before the answer ships; invented APIs are blocked; full depth on Python with existence, arity, and type; call/member/module-level grounding on 11 more languages: Java, TypeScript, JavaScript, Rust, Kotlin, PHP, Scala, C#, Swift, Go, and Ruby; 11 further languages are indexed and navigable without an answer-gate; 23 languages total, tree-sitter parsed); a merge gate, Estelle Review (grounding + secret detection + static analysis + known-CVE checks compose into one pass-or-block verdict, postable as a required GitHub check); grounded recall cited to file and line, never metered; context survival (full history stored, each step gets a window-sized cited slice, so the model's window cannot overflow); tests and repair on the team's actual suite (it keeps repairing while the suite is red and only stops at green; "no tests ran" never counts as green); the self-correcting suite, Estelle Repair + Estelle Prod, on Ultra and Team (verified fixes and proactive gap-finding; the default output is a reviewable, gate-verified PR a human merges, never an autonomous merge or deploy; auto-merge is opt-in, proof-gated, and default off; auto-deploy is designed, not shipped); change impact certified to still ground and type-check; deep search, Estelle Research, with every claim cited; an agent swarm in isolated worktrees under a configurable autonomy ceiling, every result re-gated; a library of 245 skills and playbooks, plus 16 core tools (261 MCP tools total), each callable by any agent you run; one shared team memory with role-based access control. It runs everywhere: MCP and the OpenAI-compatible API, a language server for every LSP editor (Neovim, Helix, Sublime, Zed, VS Code), and the %estelle magic in any IPython notebook (Jupyter, Colab, Kaggle).

Verified numbers, each traceable to a run log in the repo: 0 invented APIs survived the grounding gate (Python-scoped) in our verification eval (20 edits on a private repo no model had seen); recall@4 97.7% on the hard whole-repo retrieval eval; a 4,000-turn session compresses to under 800 tokens with every decision intact; LoCoMo end-to-end memory accuracy 41.7% (mid-pack, not #1; the reproducible claim is the +36pt uplift from our own baseline, not the absolute rank); self-dogfooded on Estelle's own 328-file engine, checked against 115 grounded invariants, catching real bugs in its own code.

Pricing: Free $0 (holds 10M memory-tokens, about a 2M-token codebase, one repo, unlimited recall, BYOK, no card) · Pro $19/mo (50M), $29/mo (100M), or $39/mo (150M) held, standard retrieval · Ultra $59/mo (250M), $89/mo (400M), or $139/mo (600M) held, reranking plus the self-correcting suite · Team $79/seat/mo (100M pooled), one shared memory, RBAC, self-correcting suite included · Memory Pack +50M/$15/mo, stackable · overflow pay-as-you-go at $0.35/1M memory-tokens. Yearly is 10x the monthly price (2 months free). Capacity is what you pay for; recall is never metered; model tokens are never marked up.

## Docs

- [API documentation](/docs): endpoints, config, curl and Python quickstarts
- [Tool setup](/setup): OpenCode, Cursor, Cline, Aider, Continue, Zed
- [Pricing](/pricing): the capacity tiers, overflow, and what is never metered
- [Create a key](/signup): free tier, no card

## Optional

- [llms-full.txt](/llms-full.txt): the comprehensive version of this file, with FAQ and provenance
- [Agent-mode page](/?mode=agent): this summary rendered as a browsable page