# GEODE

GEODE v0.99.87. A self-evolving autonomous execution agent.

Last sync: 2026-05-29

## Top-level pages

- Portfolio  https://mangowhoiscloud.github.io/geode/portfolio
- Docs       https://mangowhoiscloud.github.io/geode/docs
- About      https://mangowhoiscloud.github.io/geode/about
- Source     https://github.com/mangowhoiscloud/geode

## Documentation

- What GEODE is  https://mangowhoiscloud.github.io/geode/docs  — An autonomous execution agent that runs an inner agentic loop and improves itself through an outer loop. What it does, who it is for, where to go next.
- How GEODE runs a task  https://mangowhoiscloud.github.io/geode/docs/overview/how-it-runs  — One request, traced end to end. The same core serves a CLI task, a gateway message, and a scheduled self-improving run.
- The 4-layer stack  https://mangowhoiscloud.github.io/geode/docs/architecture/overview  — Model, Runtime, Harness, Agent. What each layer owns, and where its responsibility ends.
- Quick Start  https://mangowhoiscloud.github.io/geode/docs/quick-start  — Install GEODE and run your first task in a few minutes.
- Pick a path  https://mangowhoiscloud.github.io/geode/docs/run/pick-path  — Subscription, API key, or free path. How to choose for your situation.
- Configure providers  https://mangowhoiscloud.github.io/geode/docs/run/providers  — Anthropic, OpenAI, Codex, GLM. Where keys go, and what fallback chain you get.
- Run as a daemon  https://mangowhoiscloud.github.io/geode/docs/run/serve  — Start, restart, and stop serve. A thin CLI talks to an IPC-served runtime.
- The two loops  https://mangowhoiscloud.github.io/geode/docs/concepts/two-loops  — The mental model the rest of the docs build on. An inner agentic loop runs a task; an outer loop tunes the system that runs tasks.
- The inner agentic loop  https://mangowhoiscloud.github.io/geode/docs/architecture/agentic-loop  — The while(tool_use) primitive. How a turn runs, and the paths that end it.
- Context assembly  https://mangowhoiscloud.github.io/geode/docs/runtime/context  — Every LLM call's context is built here. Memory tiers plus prompt layers, under a token budget.
- Memory tiers  https://mangowhoiscloud.github.io/geode/docs/runtime/memory/5-tier  — From a raw session log to a single LLM-ready summary. Hierarchical override, budget-aware compression.
- Prompt assembly  https://mangowhoiscloud.github.io/geode/docs/runtime/llm/prompt-system  — How the system prompt is layered and assembled before each call.
- Tools and toolsets  https://mangowhoiscloud.github.io/geode/docs/runtime/tools/protocol  — The tool registry and deferred loading. A few tools load up front, the rest are fetched on demand.
- Hooks and observability  https://mangowhoiscloud.github.io/geode/docs/harness/hooks  — Lifecycle events that handlers subscribe to. How observe, react, decide, and act stack on one event.
- LLM routing  https://mangowhoiscloud.github.io/geode/docs/runtime/llm/providers  — Provider selection and fallback. Circuit breakers when a provider is unhealthy.
- Sub-agent orchestration  https://mangowhoiscloud.github.io/geode/docs/runtime/orchestration  — Spawning isolated sub-agents in parallel lanes, with snapshotted memory and merge on completion.
- Skills  https://mangowhoiscloud.github.io/geode/docs/runtime/skills  — User-invocable skills, distinct from tools. Discovery, lifecycle, override.
- The closed loop  https://mangowhoiscloud.github.io/geode/docs/capabilities/autoresearch  — The outer loop end to end. Mutate a policy, audit with Petri, attribute the result, then promote or revert.
- Co-scientist seed generation  https://mangowhoiscloud.github.io/geode/docs/capabilities/co-scientist  — A multi-role agent loop that grows the evaluation seed corpus. Supervisor, generator, critic, pilot, ranker, evolver, meta-reviewer.
- Seed pipeline  https://mangowhoiscloud.github.io/geode/docs/capabilities/seed-pipeline  — The plugin that regenerates the seed corpus each generation. Picker, orchestrator, manifest, cost preview.
- Outer-loop configuration  https://mangowhoiscloud.github.io/geode/docs/capabilities/outer-loop  — The shared schema and loader for autoresearch, seed pipeline, and Petri thresholds. Strict by default.
- Petri × GEODE  https://mangowhoiscloud.github.io/geode/docs/petri/overview  — Anthropic Alignment Science's evaluation framework, wrapped over the GEODE agent as the loop's measurement.
- Scenarios  https://mangowhoiscloud.github.io/geode/docs/petri/scenarios  — The Petri seed corpus plus GEODE-specific seeds, grouped into critical, auxiliary, and info dimension buckets.
- Run an audit  https://mangowhoiscloud.github.io/geode/docs/petri/run  — geode audit, or inspect eval for the raw path. Choose model roles, dimension set, seeds, and turn budget.
- Judge dimensions  https://mangowhoiscloud.github.io/geode/docs/petri/judge-dimensions  — What each dimension scores, and how to read the heatmap. Critical floors versus auxiliary drift.
- Seed-generation runs  https://mangowhoiscloud.github.io/geode/docs/petri/seeds  — The per-generation dashboard. Survivors, tokens, meta-review, and next-generation priors per run.
- Lineage and positioning  https://mangowhoiscloud.github.io/geode/docs/capabilities/lineage  — Where this loop sits in the self-evolving agents literature. An honest recombination of known parts, not a new primitive.
- Bundle viewer  https://mangowhoiscloud.github.io/geode/docs/petri/bundle  — The live inspect_ai transcript viewer for the latest audit run.
- Serve and gateway  https://mangowhoiscloud.github.io/geode/docs/harness/serve-gateway  — Operating the serve daemon's messaging gateway. Pollers, binding routing, lane queue.
- Messaging integrations  https://mangowhoiscloud.github.io/geode/docs/run/messaging  — Connect GEODE to Slack, Discord, or Telegram through gateway adapters.
- Schedule tasks  https://mangowhoiscloud.github.io/geode/docs/run/schedule  — Natural language and cron, with jitter. A daily report as a single command.
- Scheduler internals  https://mangowhoiscloud.github.io/geode/docs/runtime/scheduler  — How scheduled jobs are parsed, persisted, and fired.
- Lifecycle  https://mangowhoiscloud.github.io/geode/docs/harness/lifecycle  — Bootstrap, serve, shutdown. The injection order, and the cold-start lazy arc.
- Long-running safety  https://mangowhoiscloud.github.io/geode/docs/ops/long-running  — Token guards, context overflow, and a sliding window. How a long run drains gracefully.
- Cost monitoring  https://mangowhoiscloud.github.io/geode/docs/ops/cost  — Per-session and per-day budgets. The usage ledger, and when to switch models.
- Troubleshooting  https://mangowhoiscloud.github.io/geode/docs/run/troubleshooting  — Common failure modes and where to look. Logs, hooks, runlog.
- Write a tool  https://mangowhoiscloud.github.io/geode/docs/guides/custom-tool  — Define a tool, register it, and gate it with a permission policy.
- Register a hook  https://mangowhoiscloud.github.io/geode/docs/guides/register-hook  — Subscribe a handler to a lifecycle event and wire it in bootstrap.
- Add an LLM adapter  https://mangowhoiscloud.github.io/geode/docs/guides/llm-adapter  — Add a provider to the router and adapter layer, with a fallback entry.
- Configure a binding  https://mangowhoiscloud.github.io/geode/docs/guides/binding  — Route a messaging channel to a session lane with its own model and policy.
- Debug a stuck run  https://mangowhoiscloud.github.io/geode/docs/guides/debug-stuck-run  — Read the transcript and runlog, find where a run stalled, and recover.
- Configuration basics  https://mangowhoiscloud.github.io/geode/docs/config/basics  — Where configuration lives, how it loads, and how overrides resolve.
- config.toml reference  https://mangowhoiscloud.github.io/geode/docs/config/reference  — Every configuration key, grouped by area, marked stable or experimental.
- Auth and OAuth  https://mangowhoiscloud.github.io/geode/docs/runtime/auth  — Profile rotation, the Anthropic ToS path, and the Codex flow.
- OAuth token rotation  https://mangowhoiscloud.github.io/geode/docs/ops/oauth  — Refresh policy and cooldown across providers.
- System prompt modes  https://mangowhoiscloud.github.io/geode/docs/runtime/llm/system-prompt-modes  — The persona opt-in and the audit-mode strip. The two ways the system prompt is reshaped.
- Prompt caching  https://mangowhoiscloud.github.io/geode/docs/runtime/llm/prompt-caching  — The static and dynamic boundary, and ephemeral cache control.
- Prompt hashing  https://mangowhoiscloud.github.io/geode/docs/runtime/llm/prompt-hashing  — Pinned prompt hashes that break the build on unintended drift.
- MCP servers  https://mangowhoiscloud.github.io/geode/docs/runtime/tools/mcp  — Connecting MCP servers, the result-size guard, and the HTML to Markdown fallback.
- CLI and slash commands  https://mangowhoiscloud.github.io/geode/docs/harness/cli  — The thin gateway and IPC, and the slash commands it accepts.
- Guardrails G1-G4  https://mangowhoiscloud.github.io/geode/docs/verification/guardrails  — Schema, range, grounding, consistency. A fail-fast ladder over LLM output.
- BiasBuster  https://mangowhoiscloud.github.io/geode/docs/verification/biasbuster  — Confirmation, recency, and anchoring detection on top of the guardrails.
- Cross-LLM verification  https://mangowhoiscloud.github.io/geode/docs/verification/cross-llm  — An independent second opinion. A different provider re-scores the verdict.
- Cause decision tree  https://mangowhoiscloud.github.io/geode/docs/verification/cause-decision-tree  — Map a failure cause to a recovery action, instead of blindly retrying.
- Observability  https://mangowhoiscloud.github.io/geode/docs/verification/observability  — The lenses on a run. Hooks, runlog, audit diagnostics, Petri.
- Automation sidecar  https://mangowhoiscloud.github.io/geode/docs/runtime/automation  — The feedback loop and model promotion, with drift detection between agent and runtime.
- Computer use  https://mangowhoiscloud.github.io/geode/docs/runtime/computer-use  — Provider-agnostic desktop automation, Anthropic and OpenAI unified.
- CLI LaTeX rendering  https://mangowhoiscloud.github.io/geode/docs/runtime/ui/cli-latex  — Rendering math in the terminal. Detection, Unicode flatten, pretty print, image fallback.
- External references  https://mangowhoiscloud.github.io/geode/docs/reference/external-references  — Frontier agent systems, design standards, and prior work GEODE cites.
- Frontier comparison  https://mangowhoiscloud.github.io/geode/docs/reference/frontier-comparison  — Where GEODE sits next to Claude Code, Codex CLI, OpenClaw, and Hermes.
- CHANGELOG  https://mangowhoiscloud.github.io/geode/docs/reference/changelog  — Full version history, synced from CHANGELOG.md on every main push.
- Petri bundle isolation  https://mangowhoiscloud.github.io/geode/docs/reference/petri-bundle-isolation  — Operator reference for the publish workflow, validator, and hygiene ratchet.
- Architecture deep-dive  https://mangowhoiscloud.github.io/geode/docs/develop/architecture  — The subsystem map and a recommended reading order, retold as data-flow traces.
- System index  https://mangowhoiscloud.github.io/geode/docs/architecture/system-index  — Every subsystem with its file path. The flat catalog.
- Why four layers  https://mangowhoiscloud.github.io/geode/docs/explanation/4-layer  — Model, Runtime, Harness, Agent. Why the boundaries fall where they do.
- Why a self-hosting harness  https://mangowhoiscloud.github.io/geode/docs/explanation/self-hosting  — The runtime and the build line share primitives. Why that mattered.
- Why ratchet discipline  https://mangowhoiscloud.github.io/geode/docs/explanation/ratchet  — Pinned prompt hashes and a staged CI. The shape that prevents drift.
- Release and PyPI lifecycle  https://mangowhoiscloud.github.io/geode/docs/ops/release-pypi-lifecycle  — The version-bump locations, the release workflow, and the rebuild cadence.
- Backlog disposal  https://mangowhoiscloud.github.io/geode/docs/ops/backlog-dispose  — Retire an idea with a paper trail instead of a silent delete.

## Authoritative sources

- CHANGELOG  https://github.com/mangowhoiscloud/geode/blob/main/CHANGELOG.md
- CLAUDE.md  https://github.com/mangowhoiscloud/geode/blob/main/CLAUDE.md
- AGENTS.md  https://github.com/mangowhoiscloud/geode/blob/main/AGENTS.md
- llms-full  https://mangowhoiscloud.github.io/geode/llms-full.txt
