agentlas-os

Paste to Install

Paste this into the LLM you are using now, such as Claude Code, Codex, Gemini CLI, Antigravity, or Cursor:

Install Agentlas OS / Hephaestus from this GitHub repo:
https://github.com/agentlas-ai/Agentlas-OS

Run the repository's canonical one-touch installer so the durable host command
adapters are installed too; do not stop after only installing the marketplace
plugin:
curl -fsSL https://raw.githubusercontent.com/agentlas-ai/Agentlas-OS/main/scripts/install-all-runtimes.sh | HEPHAESTUS_INSTALL_GLOBAL_ROUTER=1 bash

For Claude Code, confirm that `hephaestus@agentlas-core-engine` is enabled at
user scope and that `~/.claude/commands/agentlas.md` exists. Do not report the
install complete unless a bare `/agentlas` command will be available in the
next session. At the end, confirm the active plugin, command surface, and global
routing status, then tell me to restart the host or run `/reload-plugins`.

Use this when you are already inside an LLM and want the Agentlas command surface active there. For direct shell commands, see the install methods below.

Agentlas Desktop in motion

Build an owned agent

Describe the work in plain language. Agentlas classifies the request, runs the interview and research gate, generates the package, verifies it, then asks whether to keep it only on this computer or save it privately in Agent Cloud for restore on another signed-in Desktop.

Agentlas Desktop →

Orchestrate a team

Combine local agents and borrowed Hub specialists into one orchestrator. Roles stay explicit while Agentlas manages routing, handoffs, and review boundaries.

Explore Agentlas OS →

Run and verify locally

Use the model account or API key you choose. Your current host performs the work under its local files, tools, credentials, permissions, and verification rules.

Read the trust model →

---

Build, Borrow, Own

An agent you create should remain an asset you can move, rather than a setting trapped in one chat, one model-vendor workspace, or one computer. Agentlas separates three jobs that ordinary agent builders blur together:

This is the public Agent Trust contract: a portable, owner-scoped, inspectable, and restorable package contract—not a claim of regulated financial or legal trust services.

Value	What Agentlas does	Entry point in an external LLM host
Build	Compiles a plain-language request into a runnable single-agent or team package with roles, tools, memory boundaries, permissions, routing, and verification contracts.	/agentlas build
Borrow	Finds public Hub specialists and brings the selected runtime bundle into your current Agentlas host. The publisher's private source work is not copied into your workspace.	/agentlas hub (Hub only) or /agentlas network (Local + Cloud + Hub)
Own	Keeps agents you create in a private, owner-scoped Agent Cloud so you can retrieve and call them again after changing models or computers.	Choose private Agent Cloud at /agentlas upload, then retrieve with /agentlas cloud

Portable package, local execution

Describe the work
  -> build a portable agent or team
  -> save it to my owner-scoped Agent Cloud
  -> install Agentlas OS and sign in on another supported host
  -> retrieve it with /agentlas cloud
  -> my chosen model and current host execute the work

Agent Cloud stores and retrieves the owner's package; it is not a hosted LLM that completes the work on the server. When you call a package, your selected model and current host runtime execute it under that host's permission and safety model. Credentials, local files, and machine-specific permissions do not travel with the package—you configure those separately on each computer.

The package contract — what every build must emit

An agent package is not a worker with capabilities. It is a method document. Three layers, and conflating them is the single most expensive mistake this project has made:

layer	what it is	who supplies it
LLM	the worker: reasoning, language, general knowledge	you (BYOM)
Runtime	the facilities: read the web, write files, run a shell, drive a browser	the host machine
Agent	the work manual: procedure, judgement rules, source priority, input/output contracts, stop conditions	the package

"Can this agent read the web?" is a category error. The runtime reads the web. The only question a package can answer is what method does it carry, and what must the machine be able to do for that method to run.

package-contract.json is the machine-readable list of artifacts every build emits, and scripts/verify-generated-package.sh <folder> is what enforces it. A build that omits a required artifact fails; it does not ship. Four of those artifacts carry the routing contract:

contracts/intake.schema.json     what a requester must hand over before work starts
contracts/output.schema.json     what the requester ends up holding
contracts/output.example.json    one real instance, validated by a JSON Schema
                                 validator at publish time — never by a model
.agentlas/brief.json             the compiled resume, schemaVersion agentlas.brief/1

Direction lives in the filename because nothing else ever marked it, and a schema whose direction must be guessed cannot be matched against a request.

Two rules bind every enum a build writes. Both were paid for in production:

1. Any enum reachable from matching must carry an escape member ("other", "unknown"). Matching one stated requirement against a 23-word closed list took a three-candidate inventory to zero on eight probes out of eight. A publisher whose real case is not on the list must still be findable.
2. Sentences stay sentences. No field that holds an author's sentence may be split into terms for matching. Shredding refusal sentences into the bare words tests and ci cut a correct agent's score to a quarter and pushed it from rank 2 to rank 24 — on a query that was literally its own job.

Vendor and MCP names live in exactly one place, host[].preferred, and that field is display-only. Roughly 99.9% of machines have no MCP servers installed, so a package wired to its author's own Slack, Notion or Jira has to stay usable by everyone else: the matcher reads host[].capability ("open the page in a real browser and read what a user would see") and every requirement states withoutIt, what the method still does on a machine that lacks the facility.

agentlas.brief/1 is written from either side — side: "offer" is the package's resume, side: "need" is the requester's work order. The same form, so the two can be compared without a translation step. Its schema is schemas/agentlas-brief.schema.json, and it is compiled, never hand-written: the upload path writes it from the files the package already ships.

field	what it holds	where it comes from
statement	the author's own sentence about the work	the routing card, carried whole
deliverables[].label / .contains	what the requester ends up holding, and what is inside it	contracts/output.schema.json — the title and its property names
deliverables[].shape	ledger / verdict / dossier / computation / blueprint / rendition / other	computed from the output schema's topology — arity and value spaces, never a title or a filename
deliverables[].rowVerdict / .verdictValues	the per-record judgement and its allowed values	copied verbatim from the publisher's own enum; there is nothing to match them against, so no value is ever rejected
obligations[].about / .stage	a fact the requester must supply, and when it comes to exist	contracts/intake.schema.json and the card's stated inputs
authority.performs / .gated	what this method does to the requester's world, and what it stops to ask about first	derived from the approval requirements, not from what the package calls itself
host[].capability / .withoutIt	what the machine must be able to do, and what still happens when it cannot	.agentlas/mcp-policy.json
provenance	per field: extracted, read, graded or absent	written by the compiler

provenance is the part that keeps the rest honest. absent is a legal value that scores zero, so a gap stays a gap. Filling gaps with something plausible is how capabilities came to equal snake_case(agent.md ## Responsibilities) in 130 of 130 packages — a field that was always full and never once informative.

The reference shape is Web_master: it ships input and output contracts as real JSON Schema, a domain catalogue its rules are checked against, and verification scripts that run. Everything above is what a build has to emit to be readable the way that package already is.

Deliberately absent: there is no field for what an agent is good at, and none for its permissions. Quality cannot be read off a package that has never run, and permissions are granted by the host after the fact — an employer hands you a computer once you are hired, so neither says anything about whether this method fits the work.

Hub and Agent Cloud are different scopes

Surface	What it contains	What it is for
Agentlas Hub	Public packages from creators and teams	Find and borrow only public specialists with /agentlas hub; /agentlas network federates Hub with Local and owner Cloud. Publish only through an explicit public-Hub choice.
My Agent Cloud	Only the signed-in owner's Cloud packages	Privately store, restore, and call packages you own with the /agentlas upload Cloud choice and /agentlas cloud.
Current host	The installed runtime, chosen model, local project, credentials, and granted permissions	Execute the selected local, Cloud, or Hub package.

---

The Standard, Not Another Engine

Agentlas is not building another graph engine. It is designing the agent standard — and the distribution network that carries it — that many graph engines and runtimes can share.

Engines keep changing. A model workspace, a CLI, an orchestration framework, a graph runner: each arrives, wins for a while, and is replaced. What has to survive that churn is the contract an agent carries with it — what it is, how it is invoked, what it may touch, what it remembers, and how its result is proven.

That contract is the product, and it is deliberately engine-neutral:

• One definition, many runtimes. The same package runs through Claude Code, Codex, Gemini, Antigravity, Cursor, or a local model, because the manifest, routing card, permissions, and memory rules travel with the agent instead of with the vendor.
• A life outside the chat. Package contracts, receipts, and verification gates outlive the session that produced them.
• A way to reach other people. Publish to the public Hub, or keep it in your owner-scoped Cloud. Either way it moves as a package, not as a copied prompt.

Hephaestus is one engine that implements this standard. It is open source precisely so the standard can be inspected, forked, and re-implemented — including by engines we did not write.

Why Agentlas OS

Most AI products help you create another agent. Agentlas OS is for the harder part: making agents operate as a team you own.

You should be able to imagine this after installing it:

• Your LLMs work like a team instead of isolated chat sessions.
• Your real browser becomes an execution surface, not a screenshot in a prompt.
• Your agents keep package contracts, routing cards, memory rules, permissions, and verification receipts after the chat ends.
• Packages you own can remain local or be privately stored in your owner-scoped Agent Cloud, then retrieved from another supported, installed, signed-in host.
• Your existing Claude Code, Codex, Gemini, Cursor, Antigravity, API keys, and local models become part of one operating layer.
• Hub specialists can be borrowed into your local runtime without copying the creator's private work or sending your private files to their agent.

Hephaestus is the open-source engine underneath Agentlas OS. It is not a prompt marketplace, an agent template generator, or another model subscription. It is a local-first runtime that builds, routes, borrows, runs, verifies, and packages agents across LLM command surfaces.

The point is not "make an agent from a prompt." The point is:

Create, package, route, run, and verify agents across your LLMs, browser, memory, and local tools.

Why Not Just Make A Claude Agent?

Claude subagents and custom agents are useful. They give a task its own prompt, tools, and context window. Agentlas starts after that point.

An LLM can draft an agent. Agentlas turns it into an operating unit:

Layer	A prompt-made agent	An Agentlas package
Definition	Role prompt, markdown, tool list	Manifest, agent card, mode map, package contract
Invocation	Manual mention or simple trigger	Routing card, triggers, anti-triggers, benchmarks, receipts
Browser	Ad hoc browsing or screenshots	Real browser hardpoint with visible clicks, forms, waits, and snapshots
Memory	Copied context or chat history	Memory map, memory tickets, Memory Curator, Policy Gate
Runtime	One LLM session or one vendor runtime	Adapters across Claude Code, Codex, Gemini, Cursor, Antigravity, and local runtime
Teams	Another prompt layer	Orchestrator, PM Soul, Memory Curator, Policy Gate, eval judge, QA gate
Verification	User checks manually	Package checks, receipts, Stormbreaker final gate
Ownership and portability	Trapped in the chat or vendor workspace where it was created	Portable package that can remain local or be retrieved from the owner's Agent Cloud on another supported, installed, signed-in host
Distribution	Copy the prompt	Explicit choice between public Hub publishing and private owner-scoped Cloud storage

That is the product boundary: Agentlas does not compete on "better prompt." It gives agents the architecture to keep working outside one chat.

The Agent OS Stack

Agentlas maps agent work to operating-system-like responsibilities without forcing your work into one model provider:

OS Abstraction	Implementation in Hephaestus
Kernel / Policy Gate	Deterministic router + security gates. Every routing action yields an auditable receipt; tool execution permissions are enforced by the active host and runtime.
Processes / Threads	Independent agents and multi-agent teams compiled as packages with explicit, typed contracts (Routing Cards, anti-scopes, memory boundaries, and verification shims).
Process Scheduler	Network 2.0 routing (local-first, quality-gated, and benchmark-gated dispatch) combined with Stormbreaker's parallel execution fabric and append-only run journals.
Memory Management (MMU)	Two-boundary governed memory: local project memory remains isolated on the machine, while durable promotions are gated by a local Memory Curator.
Virtual File System	Production Ontology Runtime: local-first source ingestion, CJK trigram FTS5 search, hybrid Reciprocal Rank Fusion, and GraphRAG retrieval.
Inter-Process Call (IPC)	A2A Agent Card Boundary (cryptographic import/export and caller-gating) + Model Context Protocol (MCP) tool registrations.
Package Manager	Agentlas Hub for public publishing and borrowing; owner-scoped Agent Cloud for private package storage and retrieval. Neither is a server-side model executor.
Shell Interface	A small, unified command set in external client runtimes; plain-language intent routing in native Agentlas shells.
Process Initialization	Meta-Agent Factory with an integrated Briefing Interview Gate—specifying agent parameters before compiling code.

---

New In v1.1.0 — The Briefing Interview Engine

Agents generated from vague, single-sentence prompts fail under real-world edge cases. Hephaestus v1.1.0 positions task specification as a first-class OS service through the Briefing Interview Engine:

• Quantitative Ambiguity Gates: The compilation scheduler evaluates prompt clarity across four key vectors (Goal, Constraints, Scope, Context). The build process is strictly gated until the ambiguity score passes a numeric threshold (ambiguity score $\le 0.2$, with per-dimension safety floors). Clear prompts bypass the interview loop entirely via a budget system that caps questions for trivial tasks.
• Lens-Driven System Analysis: Clarifying questions are dynamically sourced from a structured lens table (Scope, Intent, Challenge, System Architecture) focusing on critical routing indicators: anti-scope bounds (what the agent must NOT do), verifiable acceptance criteria, and exit conditions.
• The Work Brief: Resolved details are frozen into .agentlas/work-brief.json—recording the validated goal, concrete constraints, an assumption ledger with source tags, and the metadata ambiguity score.
• Contextual In-Flight Briefs: The CLI tool cards migrate automatically maps brief details directly to triggers and anti-triggers on the agent's routing card. Running route --brief propagates this brief to all Stormbreaker execution packets, ensuring constraints and exit conditions govern parallel subprocesses across the entire lifecycle.
• Enhanced Routing Discrimination: Prevents same-topic/different-intent collision (e.g., a security agent intercepting a deployment prompt) via double-sided gating: interview-validated anti-triggers on the routing card, and low-confidence LLM re-ranking escalation inside the router.

---

All Install Methods

Manual LLM Adapter Install

Use this only when your current LLM cannot run setup for you. It installs the shared Hephaestus runner and command adapters for supported LLM tools.

xcode-select --install   # Command line tools (skip if already installed)
git --version            # Confirm git is available
curl -fsSL https://raw.githubusercontent.com/agentlas-ai/Agentlas-OS/main/scripts/install-all-runtimes.sh | bash

This installs the neutral runner at ~/.agentlas/runtime/current/bin/hephaestus and registers the command adapters for Claude Code, Codex, Gemini CLI, Antigravity, Cursor, OpenCode, OpenClaw, Hermes, and compatible local/API hosts. The installer verifies each runtime surface after registration.

Desktop startup and every /hep-* command start the same digest-verified, rate-limited update check in the background. The current command never waits for network or installation work. A successful update atomically moves ~/.agentlas/runtime/current and reconciles every already-installed host adapter; the next command or reloaded session uses the new release. Releases v1.1.63 through v1.1.68 briefly installed a separate six-hour OS scheduler. Current installs remove that retired scheduler automatically.

Optional Global Router

hephaestus global install

This appends a managed marker block to ~/.codex/AGENTS.md, ~/.claude/CLAUDE.md, and ~/.gemini/GEMINI.md. It is an optional host adapter, not the owner of an Agentlas One session or a Desktop Work project. For substantial work, Network is one explicit federated source scope (local + cloud + hub): the host model selects exact releases from the returned menu and validates them. It never routes by keyword, silently substitutes a different agent, or treats Cloud, Hub, Local, and skills as semantic fallback tiers. The command is idempotent and keeps a timestamped backup before editing.

The installed router prompt names final workers, not router commands. It carries an explicit status-line contract for English and Korean sessions:

Session language	Agent route example	Host-skill adapter example
English	Agents used: <agent names>. Reason: <short reason>.	Skills used: <skill names>. Reason: <short reason>.
Korean	사용 에이전트: <agent names>. 이유: <short reason>.	사용 스킬: <skill names>. 이유: <short reason>.

Global router command reference:

Command	What it does
hephaestus global install	Install or refresh the managed router block for Codex, Claude Code, and Antigravity/Gemini.
hephaestus global status	Show whether each runtime file has the managed router block.
hephaestus global remove	Remove only the managed Hephaestus router block. Existing user content stays in place.
hephaestus global install --target codex	Install only ~/.codex/AGENTS.md.
hephaestus global install --target claude	Install only ~/.claude/CLAUDE.md.
hephaestus global install --target antigravity	Install only ~/.gemini/GEMINI.md, which Antigravity shares with Gemini CLI.
hephaestus global install --target codex --target claude --target antigravity	Explicitly install all supported targets.
hephaestus global install --dry-run	Preview what would change without writing files.
hephaestus global install --no-backup	Edit without writing a timestamped .bak.* file.
hephaestus global install --home /tmp/test-home	Test against another home directory. Useful for installer QA.
hep-global install	The original spelling of the same command. Still supported.
~/.agentlas/runtime/current/bin/hephaestus global status	Use the installed runtime directly when shell shims are not on PATH.

The independent Agentlas Terminal owns the agentlas shell command. The Core installer intentionally leaves that command untouched and removes only the exact legacy Core alias it created in older releases.

Per-Runtime Plugin Drivers

From your OS terminal:

claude plugin marketplace add https://github.com/agentlas-ai/Agentlas-OS --sparse .claude-plugin claude/plugins
claude plugin install hephaestus@agentlas-core-engine

Note: Claude Code always namespaces marketplace-plugin commands, so this plugin-only path exposes /hephaestus:agentlas. To get the documented bare /agentlas autocomplete in every new session, use the one-touch installer above; it also writes ~/.claude/commands/agentlas.md and hep-*.md. Claude Code supports claude plugins ... as an alias, but this README uses the singular claude plugin ... for consistency.

From your OS terminal:

codex plugin marketplace add agentlas-ai/Agentlas-OS --ref v1.1.105
codex plugin add hephaestus@agentlas-core-engine

Note: Codex does not accept /plugin marketplace add inside the app — run the two commands above in your OS terminal. The OS-terminal CLI command is singular (codex plugin); inside the Codex app, the plugin browser slash command is plural (/plugins). Codex 0.117+ removed custom /prompts:* commands; after install, invoke the supported plugin skill as $hephaestus-network <request>.

Clone the repo and copy AGENTS.md, agent.md, agents/, skills/, modes/, schemas/, templates/, and .agentlas/ into your workspace. Runtime folders (.claude/, codex/, .gemini/, .agents/) function as adapters over the same canonical core.

Just talk: After installation, speak in plain language within native Agentlas interfaces to auto-route tasks. In external LLM tools, use the explicit commands listed below. When you don't know what agents exist, start with /agentlas search. To connect Telegram, use /agentlas connect.

One word to remember: everything starts with agentlas — /agentlas in an LLM host, agentlas in a shell.

---

Where This Fits

This repository installs the Hephaestus engine and LLM command adapters. It is the open-source command surface under Agentlas OS.

Surface	Role
Agentlas Desktop	Visual local OS for running AI-native apps, agent teams, memory, browser work, and Hub specialists.
Hephaestus plugin	Open-source engine and command surface for Claude Code, Codex, Gemini CLI, Antigravity, Cursor, and compatible runtimes.
Agentlas Hub	Public package surface for publishing and borrowing specialists.
Agentlas Cloud	Owner-scoped package store for privately saving and retrieving the signed-in user's own agents.

The install prompt above is intentionally scoped to this repo and the current LLM surface. Desktop, Hub, and Cloud are product surfaces around the same Agentlas OS architecture; they are not prerequisites for installing the plugin. Cloud retrieval on a new computer does require a supported Agentlas OS host to be installed and the package owner to be signed in.

---

The Command Surface

Inside native Agentlas environments, Hephaestus operates commandless. External LLM tools use a deliberately small visible command set. System-level utilities like Stormbreaker, research loadouts, and configuration tables attach automatically from context:

System Subsystem	Command	Example
Agent / Team Builder	/agentlas build	/agentlas build create a customer support agent for Shopify refunds
Workforce Federation (Local + Cloud + Hub)	/agentlas network	/agentlas network split this launch plan into research, copy, QA, and release agents
Registered Local Agents Only	/agentlas local	/agentlas local use only agents registered on this machine
Owned Cloud Agents Only	/agentlas cloud	/agentlas cloud use my saved finance analyst agent to review this report
Public Hub Agents Only	/agentlas hub	/agentlas hub find public specialists for accessibility QA
Directory Search	/agentlas search	/agentlas search find agents for a market report workflow
Browser Hardpoint	/agentlas browser	/agentlas browser https://example.com
Inter-Process Call (IPC)	/agentlas call	/agentlas call market-researcher, report-writer {draft a market report}
Cloud / Hub Destination Gate	/agentlas upload	/agentlas upload ./agents/customer-support-hq
Telegram Setup	/agentlas connect	/agentlas connect Telegram for Marketing Agent Team

Every row also answers to its original /hep-* name — /agentlas network and /hep-network are the same command. Nothing was renamed away, so existing scripts, notes, and muscle memory keep working. Current Codex uses explicit plugin skills such as $hephaestus-network, $hephaestus-build, $hephaestus-cloud, and $hephaestus-storm; other MCP surfaces can be requested in plain language. Custom /prompts:* commands were removed from Codex 0.117+.

---

The OS Subsystems

Meta-Agent Factory — Process Creation

A unified compilation factory using three builders. Every generated package registers its global command (.agentlas/global-commands.json) and ships verification scripts—the user never has to infer how to run the compiled package:

Compilation Mode	Routing Target	Output Artifact
Single-Agent	10-single-agent-builder	Standalone worker with localized skills, memory contracts, and runtime adapters.
Multi-Agent Team	20-multi-agent-team-builder	Hierarchical team containing a PM Orchestrator, Memory Curator, Policy Gate, QA, and validation scripts.
Workspace Packager	30-agentlas-packager	Compiled bundle ready for runtime import, CLI execution, or GitHub distribution.

Briefing Interview Gate: Builders initiate the process using the briefing interview gate (docs/builder-interview-research-gate.md): conducting lens-driven questions, evaluating the ambiguity threshold, searching primary sources, and outputting the work brief.

---

Network 2.0 — The Scheduler

Figure 2. A2A scheduling: explicit host commands, model-selected exact releases, local memory, and federated Agentlas Network discovery.

• Typed Job Analysis: The active host LLM turns the request into a redacted WorkOrder with explicit roles, skills, tools, artifacts, authority, cardinality, and handoffs. Core does not infer staffing intent from a substring list.
• Exact Source Federation: local, cloud, and hub are exact scopes; network is their sealed union. Each source returns a bounded, content-only menu, and Core records unavailable sources instead of silently widening scope.
• Host-Owned Task Forces: The host LLM reads the qualification evidence and authors the exact Selection. Core never chooses a deterministic winner or performs a hidden Router Agent re-rank; it validates governance, privacy, identity, cardinality, and graph integrity.
• Pinned Execution: After validation, Core fetches only the selected immutable releases from their original source sessions and verifies release, package, and content digests before distinct planner, worker, synthesis, and verifier invocations run.
• Evidence-Scored Evaluation: Retrieval coverage, host selection, immutable preparation, real child invocations, and final verification are scored as separate claims. A benchmark result never replaces the host staffing decision or turns usage history into routing authority.

Details: docs/hephaestus-network-2.0.md · Runtime support matrix: docs/runtime-fallback-adapters.md

---

Stormbreaker — Disciplined Execution

Stormbreaker is the execution gating subsystem of the Agent OS. It ensures that agents do not report success or terminate until all outcomes have been verified by deterministic checks:

Kernel Gating Envelope:
[Scope Lock] -> [Decomposition] -> [Parallel Work Packets] -> [Verify Contracts] -> [Bounded Repair] -> [Final Gate]

A local run journal makes long executions resumable after interruption. Execution packets carry the Work Brief so that anti-scope rules and exit criteria govern all parallel subprocesses. Stormbreaker reports explicit completion states (verified / unverified / blocked) to prevent autonomous completion theater.

Execution protocol: docs/robustness-protocol.md · Benchmarks & Evals: docs/robustness-eval.md

---

Ontology Runtime — The Knowledge Filesystem

For knowledge-intensive operations, bin/ontology acts as the semantic filesystem, converting unstructured local files into an agent-readable database stack:

Ingested Files -> [Parser Adapter] -> [CJK trigram/bigram tokenization] 
  -> [FTS5 + SQLite Storage] -> [Reciprocal Rank Fusion Ranking] -> [GraphRAG Search]

Features first-party Korean document parsing (HWPX and legacy HWP5) with zero GPL dependencies. Fully local and SQLite-backed; confidential and private chunks are isolated, preventing them from reaching external cloud hooks.

The v1.1.56 release contract ships and verifies a dependency-free potion-base-8M int8 Model2Vec asset as the primary semantic adapter. Its normalized 256-dimensional semantic vector is combined with a normalized hash-96 vector into one fixed 352-dimensional local vector. Runtime queries never download a model or call a hosted embedding API. Hash-only mode is an explicitly reported degraded fallback when the verified local asset is missing or rejected, not an alternative silent default.

The v1.1.56 self-updater installs the complete one-touch runtime payload, including Career Graph, templates, and the verified model under the versioned models/model2vec/potion-base-8M-int8 directory. It checks that payload before and after switching ~/.agentlas/runtime/current, then repairs merge-safe memory hooks for detected hosts without replacing unrelated user configuration.

Agent experience recall is a governed path, not an unrestricted nearest-vector search:

exact agent + allowed scope + active status + unexpired + not superseded
  -> lexical rank + local cosine rank
  -> reciprocal-rank fusion + bounded salience prior
  -> all relevant memories when they fit, otherwise budgeted top-k

Every governance-eligible experience row is considered before token-budget selection, so an arbitrary recency window cannot hide older evidence. Each Hub agent has a rebuildable private projection at ~/.agentlas/networking/hub-agents/<normalized-slug>/memory/experience.sqlite. The runtime may infer only same-agent, same-scope similar_to edges from local cosine similarity; supersedes and contradicts require an explicit curator decision.

bin/ontology ingest ./corpus --scope internal
bin/ontology --db .agentlas/ontology-runtime.sqlite query "Project Helios Memory Curator" --agent verifier
bin/ontology --db ~/.agentlas/networking/hub-agents/<slug>/memory/experience.sqlite experience query "What did we learn?" --agent hub:<slug>
bin/ontology memory candidates

Plain Claude Code and Codex sessions receive bounded recall through SessionStart and UserPromptSubmit additional context. Antigravity uses a PreInvocation ephemeral message, OpenCode uses an experimental local plugin, and Grok refreshes a workspace-scoped cache because its passive hooks do not inject stdout. These hooks supplement live AGENTS.md/CLAUDE.md policy rather than copying it. Details: docs/ontology-runtime.md · docs/runtime-memory-hooks.md

---

Governed Memory — Curated Promotion

• Local Project Memory: Project documents remain in the local .agentlas/ontology-runtime.sqlite; borrowed-agent experience remains in its exact per-agent projection. The two stores share one query engine without collapsing their scope or ownership boundaries.
• Governance Before Ranking: Exact agent, allowed privacy scope, active status, expiry, and structural supersession are enforced before lexical/cosine ranking. Secret redaction and capsule bounds are applied again before host delivery.
• Workspace Personalization: Manages summaries, playbooks, plugin locks, and receipts for borrowed Cloud/Hub agents without storing raw prompts, credential values, or private files.
• Curator Gating: Skills and durable memory modifications remain candidates until a local curator confirms evidence, rollback coverage, and security policy approval. Automatic experience relations are limited to similar_to.

---

A2A Boundary — Inter-Agent Isolation

Standardized CLI commands allow safe inter-agent coordination:

agentlas-cloud ao a2a import ./agent-card.json .
agentlas-cloud ao a2a export . --agent local/10-builder
agentlas-cloud route "run the release check" --caller local/orchestrator .

Import acts as a proposal (restricting automatic invocation), export redacts private paths and logic, and invocations are caller-gated before routing is resolved.

---

Built For Owned Agent Operations

Users and teams do not need another way to write isolated agents. They need to operate an owned workforce of them. Hephaestus is designed for that operational model:

• Package Portability: Packages you own can remain local or be stored in your private, owner-scoped Agent Cloud. On another supported computer, install Agentlas OS, sign in, and retrieve the package; host credentials, local files, and permissions stay machine-specific.
• Model Neutrality: Agent packages use adapters for supported hosts instead of belonging to one model vendor's workspace. You can run the same package with supported Claude, Codex, Gemini, Antigravity, Cursor, or local-model surfaces without rebuilding its operating architecture.
• Auditability by Construction: Every routing decision, execution step, memory candidate, and curator decision is logged as a text file. You can diff, audit, and commit them. Work is either verified or flagged as unverified.
• Deterministic Pipeline Gates: Security filters, anti-scopes, routing card triggers, and prompt sanitizations are hardcoded into the OS pipeline—they do not rely on LLM system instructions or guidelines.
• Specification Before Generation: The Briefing Interview Engine measures request ambiguity and stamps the score on the Work Brief, ensuring task execution can always be audited back to what was agreed.
• Local-First Data Boundary: Raw text, documents, and database files remain local. External transactions are redacted and opt-in.

Where Frameworks Fit

CrewAI, LangChain, and vendor agent SDKs function as libraries—excellent for writing custom agent logic inside a single process. Hephaestus operates as the runtime substrate: it specifies, packages, routes, runs, audits, and migrates agents across workspace runtimes. Framework code runs inside Hephaestus packages; the kernel only requires that agents honor their directory contracts and Routing Cards.

---

What It Builds (Process Packaging)

Hephaestus packages agents into a standard directory layout that any workspace runtime can parse, install, verify, and run. The important part is not just agent.md; it is the operating contract around it:

├── AGENTS.md                              # Canonical operating loop and source-of-truth map
├── agent.md / agents/                     # Single worker, HQ/orchestrator, or team roles
│   ├── 10-single-agent-builder/
│   ├── 20-multi-agent-team-builder/
│   └── 30-agentlas-packager/
├── .agentlas/                             # Agentlas OS system directory
│   ├── sitemap.json                       # Product graph: modes, runtime adapters, memory, release checks
│   ├── mode-map.json                      # Single-agent / team / packager classification contract
│   ├── routing-card.json                  # Triggers, anti-triggers, capabilities, risk, routing readiness
│   ├── agent-card.json                    # A2A-facing identity and capability card
│   ├── company-blueprint.json             # Team/company topology for multi-agent packages
│   ├── global-commands.json               # Runtime command aliases and install surfaces
│   ├── memory-map.json                    # Memory roots, write owners, trust labels, exclusions
│   ├── memory-tickets.jsonl               # Candidate memory events before durable promotion
│   ├── project-soul-memory.md             # Project-level operating memory
│   ├── curator-decisions.jsonl            # Memory Curator promotion/rejection decisions
│   ├── vault-references.json              # Secret/credential references without raw values
│   ├── validation-ledger.jsonl            # Verification and release evidence
│   ├── field-test-report.json             # Field test results for package readiness
│   ├── skill-registry.json                # Reusable skill inventory and lifecycle metadata
│   ├── skill-trials.jsonl                 # Skill trial evidence before promotion
│   ├── agent-ontology/                    # Local code/agent map for capabilities, artifacts, scopes, edges
│   └── super-ontology-*.json/jsonl        # Governance contracts: evidence, privacy, side effects, resilience
├── skills/                                # Canonical reusable skills
├── modes/                                 # Mode contracts for build/package behavior
├── schemas/                               # JSON schemas for cards, memory maps, sitemap, evals, manifests
├── templates/                             # Package, memory, interview, eval, ontology, and contract templates
├── ontology/ + bin/ontology               # Local-first parser/search/GraphRAG runtime
├── agentlas_cloud/                        # Hub/Cloud bundle, routing, update, and runtime APIs
├── .claude/ codex/ .gemini/ .agents/      # Thin runtime adapters over the same core
├── claude/ codex/ gemini/ antigravity/    # Plugin/extension/workflow distributions
├── cursor/ hermes/ openclaw/              # Additional runtime shims and skill mirrors
├── docs/                                  # Architecture, chain map, memory, ontology, routing, eval docs
│   ├── source-of-truth.md
│   ├── chain-map.md
│   ├── memory-architecture.md
│   ├── ontology-runtime.md
│   ├── hephaestus-network-2.0.md
│   └── builder-interview-research-gate.md
└── scripts/                               # Verification, installer, sync, release, and public-safety gates
    ├── verify-package.sh
    ├── verify-ontology-runtime.sh
    ├── verify-routing-cards.sh
    ├── sync-adapters.sh
    └── public_safety_check.sh

That package shape is why an Agentlas agent is more than an LLM-written role prompt. It carries routing, memory, sitemap, code/agent ontology, permissions, runtime adapters, verification ledgers, and release gates together.

Dependency-aware project context

Hephaestus keeps the project map local, refreshes it by source fingerprint, and hands each concrete task only the goals, constraints, definitions, backlinks, interfaces, and related files it structurally depends on.

Code Map v2 also carries an agentlas.verification-map.v1 graph. It links source files to the tests that verify them, package test commands, CI workflows, and product-version contracts. Those non-source files participate in the map fingerprint, so changing a workflow or version manifest invalidates the same snapshot as changing code. context impact follows these links, and context verify blocks completion until affected source dependents are changed or reviewed and one linked execution channel is satisfied: run the local tests, run the CI workflow, or run both. Local and CI are alternatives, not duplicate requirements. Version contracts remain a separate release responsibility, and a missing test reference blocks the CI channel that owns it until the workflow is fixed or explicitly waived. Local test files and fixtures may remain Git-ignored and excluded from public runtime archives; the project-local map still fingerprints and links them so local execution is a real selectable channel rather than invisible state.

hephaestus context refresh --project .
hephaestus context refs resolveHubEntityKind --project .
hephaestus context slice --project . --task "fix entity kind routing" \
  --target src/package-kind.ts --render
hephaestus context impact --project . --changed src/package-kind.ts
hephaestus context verify --project . --changed src/package-kind.ts \
  --reviewed src/register/route.ts

Claude and Codex receive the task slice from their local hook and get a reverse-reference warning immediately before edit tools. Desktop, Terminal, Stormbreaker, and Workforce call the same Core implementation. Network/Cloud search and bundle fetches remain redacted and never receive local source paths or Context Map contents.

---

Docs By Goal

System Goal	Reference Documentation
Understand the canonical route	AGENTS.md
See the full team contract	agent.md
Architecture source of truth	docs/source-of-truth.md
Runtime boundaries	docs/runtime-sync-boundaries.md
Sitemap contract	.agentlas/sitemap.json and schemas/sitemap.schema.json
Mode map	.agentlas/mode-map.json
Routing card	.agentlas/routing-card.json and schemas/routing-card.schema.json
Memory map	.agentlas/memory-map.json and schemas/memory-map.schema.json
Briefing interview & research gate	docs/builder-interview-research-gate.md
Network 2.0 routing	docs/hephaestus-network-2.0.md
Stormbreaker protocol	docs/robustness-protocol.md
Canonical Goal + UltraCode harness	docs/stormbreaker-goal-ultracode-harness.md
Ontology runtime	docs/ontology-runtime.md
Memory architecture	docs/memory-architecture.md
Runtime memory hooks	docs/runtime-memory-hooks.md
Experience and Taste assets	docs/agent-experience-assets.md
MCP build resolution	docs/mcp-build-resolution.md
Plugin contributions	PLUGIN_CONTRIBUTIONS.md
Model allocation	docs/model-allocation.md
Skill lifecycle promotion	docs/skill-lifecycle-promotion.md
Cloud runtime bundles	docs/agentlas-cloud-runtime.md
Verify a package	scripts/verify-package.sh
Public safety check	scripts/public_safety_check.sh

---

Public Safety Boundary

This repository does not include Agentlas billing/account logic, production cloud credentials, customer databases, raw private transcripts, native keychain managers, or private deployment scripts.

Public output packages compiled by Hephaestus must exclude local absolute paths, API keys, service-account keys, .env secrets, raw transcripts, customer logs, or private developer notes.

---

Contributing and Verification

Read CONTRIBUTING.md before opening a pull request. External services and optional providers must be submitted as independently installable Agentlas plugins; provider-specific Core wiring is not accepted. The detailed boundary and review checklist are in the plugin contribution guide.

Before opening a pull request or publishing updates, run the verification test suite:

scripts/verify-package.sh
scripts/verify-ontology-runtime.sh
scripts/verify-experience-assets-contract.sh
scripts/public_safety_check.sh

---

License

Apache-2.0. See LICENSE.