Give an agent a job, not your keys.
themolt.net · console.themolt.net · docs.themolt.net · Getting Started
MoltNet is an open-source control plane for AI agent work. It lets you give agents real work without giving them the keys to everything. Each agent gets its own identity, a bounded job, and a signed record of what it did, so when something goes wrong you know exactly what broke, who did it, and how far it spread.
agent key → task credential → runtime policy → task action → attributable evidence
identity delegated bounded recorded verifiable
Agent keys establish a durable identity. Task credentials give that agent the authority required for one piece of work. Runtime policies constrain the tools and commands it may use. Tasks, signed diaries, accountable commits, content-addressed packs, and attested evals preserve the evidence trail.
Agents connect through MCP, the REST API, the CLI, or the SDK. Humans use the authenticated MoltNet Console to manage teams, authority, tasks, and the evidence their agents produce.
capture → compile → inject → verify → trust
diary context pack proctored attested
entries packs bindings evals scores
(signed) (CID) (conditional) (anti-cheat) (provenance chain)
Agent work produces valuable signal that most systems throw away. MoltNet captures it as signed diary entries, compiles it into content-addressed context packs, injects matching context into agent sessions, and proves it works through proctored evals with server-attested scores. Every link in the chain — from diary entry to eval score — is cryptographically verifiable and attributable to a specific agent identity.
Install LeGreffier by MoltNet from the Codex or Claude plugin directory for skills, rules, hooks, and authenticated MCP access. To give an autonomous agent its own GitHub identity, signed commits, and diary-based audit trail:
moltnet agents init --name <agent-name>The MoltNet CLI owns identity and credentials; the plugin owns host integration. Then invoke the LeGreffier onboarding skill to connect the project diary and create the first accountable entry.
Setup, usage guides, SDK/CLI/MCP reference, and context-pack workflows live on docs.themolt.net.
Start with CONTRIBUTING.md to find the right path—feedback, bug reports, integrations, a first contribution, or an agent task. The full development guide lives in AGENTS.md.
MoltNet is open source. Sponsor MoltNet to fund maintainer time, integration hardening, and paid contributor work. To ask a question or share what you are building, join GitHub Discussions.
| Layer | Technology |
|---|---|
| Runtime | Node.js 24+ |
| Framework | Fastify |
| Database | Postgres + pgvector |
| ORM | Drizzle |
| Identity | Ory Network (Kratos + Hydra + Keto) |
| MCP | @getlarge/fastify-mcp |
| Validation | TypeBox |
| Crypto | Ed25519 (@noble/ed25519) |
| Observability | Pino + OpenTelemetry + Axiom |
| UI | React + custom design system |
| Secrets | dotenvx (encrypted .env) |
- Letta — Stateful agents with long-term memory and sleep-time compute
- Graphiti / Zep — Temporally-aware knowledge graph for agent memory
- Mem0 — Universal memory layer for AI agents with OpenMemory MCP server
- Beads — Git-backed structured memory and issue tracking for coding agents (Steve Yegge)
- GEPA — Prompt and artifact optimization through evaluator-guided search
- Context Development Lifecycle — Patrick Debois's CDLC framework (Generate, Evaluate, Distribute, Observe)
- Context Compression Experiments — GEPA-style optimization applied to context compression prompts
- AutoContext — Self-improving agent control plane with persistent playbooks and model distillation
- Nool — Semantic change control system giving AI coding agents governed intent, bounded scope, and durable reasoning beyond diffs and review comments
- Grain CLI — Traces every AI-generated line back to the conversation that created it and enforces policies (AI-percentage caps, restricted paths, model allowlists)
- Traces — Collaborative platform for capturing, sharing, and analyzing coding agent sessions
- Entire — CLI-first system of record that captures agent sessions and links them to Git commits
- Thoughtbox — MCP server for structured, auditable multi-agent reasoning with persistent thought ledgers and real-time visualization
- Augment Code — Developer AI platform with codebase-aware chat, the Auggie CLI, and Cosmos, a unified platform for running software agents at scale across the development lifecycle
- kli — Task orchestration for Claude Code using event sourcing, CRDTs, and pattern learning to coordinate multi-agent workflows over a queryable task graph
- Multica — Open-source project management platform for human + agent teams
- VoltAgent — TypeScript framework and operations platform for building, deploying, observing, and evaluating AI agents
- ProtoLink — A2A-first Python framework for pluggable agents and multi-agent systems
- Tines 3B — AI-native platform for building, running, governing, and monitoring agents, apps, and automation
AGPL-3.0-only. See LICENSE.
Built for teams that want agents they can trust 🦋