The production Access Engine for Hermes Agent.

World Model · Workbench · Quick Start · CLI · Docs · About

Hermespace gives Hermes Agent a bounded, session-safe working room. It selects active concepts before a material turn, keeps silent intermediates out of user chat, observes native tools/subagents, and preserves continuity after the turn. Hermes remains the actor; Hermespace keeps the problem coherent.

Install Space and any Hermes Agent gets a better turn: a bounded desk, plus optional Cube library and Insight pattern card as organs behind one unioned plugins.enabled. Cube and Insight stay standalone repos — Space soft-imports them when present and never rewrites a plugin list or a memory provider you already chose.

Production target: Hermes Agent v0.20.0+ — CLI, gateways, A2A, tools, subagents, turn boundaries, and finalization.

---

The World Model

A persistent agent world. Standalone, the local JSONL warehouse may grow. When HermesCube is installed, World projects from the Cube book and does not grow a second forever-archive.

Archive

from hermespace import WorldModel

wm = WorldModel(agent_id="my-agent")
wm.enter()                                  # agent enters the world
wm.add_belief("Fast deploys reduce risk", 0.85, source="observation")
wm.add_belief("Fast deploys reduce risk", 0.85, source="observation")  # reinforces
wm.add_landmark("Zero-downtime deploy v2")
wm.set_trait("operations-minded")
wm.set_goal("Improve deploy pipeline", decision="A — automate", plan=["audit", "implement", "verify"])
wm.add_relationship("dev-team", "team", 0.8)
print(wm.render_markdown())                 # epoch-aware markdown
wm.evolve()                                 # consolidate, detect patterns, check epoch

Standalone mutations append to ~/.hermespace/worlds/{agent_id}_archive.jsonl. With Cube, World is a projection (pulse_charge / sync_world_beliefs) and that JSONL is not a second book.

Entry types: enter, leave, landmark, belief, trait, evolution, focus, epoch_transition, resolve, relationship.

Epochs

The agent's stage is measured by accumulated experience, not wall clock:

Epoch	Entries	Behavior
Genesis ✨	<5	Brief timeline, all entries visible
Growth 🌱	5–24	Section headings, summarized older entries
Maturity 🎓	25–99	Condensed timeline, wisdom from top beliefs
Wisdom 🧠	100+	Evidence-gated wisdom, deeply held beliefs only

Epoch transitions are recorded as epoch_transition archive entries and change how the world is rendered in the agent's context block.

Beliefs

Beliefs have confidence (0–1) and corroboration count. Calling add_belief with the same statement reinforces — confidence increases by +0.1, corroboration increments. The evolve() cycle consolidates semantic notes from SemanticStore and detects recurring patterns in landmarks.

Causal Chains

Every entry can reference parent entries via causal_parents. set_goal() links to the most recent timeline entry. trace_chain() walks backward through the archive.

# Walk the chain
focus_entries = wm.archive.query("focus")
chain = wm.trace_chain(focus_entries[-1].id)

# Resolve an outcome
wm.resolve_outcome(lm_entry.id, "success")

Evolution

The evolve() cycle runs five stages:

1. Consolidate semantic notes into beliefs
2. Detect recurring patterns in recent landmarks
3. Analyze relationship affinity trends
4. Detect milestones (entry count thresholds, high-confidence beliefs)
5. Generate open questions from low-confidence beliefs

Then checks for epoch transition, refreshes active concepts (Access Workspace hub, max 25), and writes an evolution archive entry.

The world_evolve pulse job runs this hourly. Manual: hs world evolve or WorldModel.evolve().

Context Injection

One user-message inject, progressive disclosure. Mid-load target ≤2.8k (hard cap <9k); high/protect ≤900. Fluent ack, high load, or a missing organ injects nothing. Cube is skipped when the provider already prefetched. Spoken Report stays short (line 1 = next action); dense context is not also dumped into chat. Harvest runs on finalize (≤10s, fail-open) and never rides the inject path.

CLI

hs world show               # render world.md
hs world enter              # agent enters the world
hs world leave              # agent leaves (session end)
hs world evolve             # run evolution cycle
hs world add-belief --statement "..."
hs world add-landmark --event "..."
hs world set-trait --trait "..."
hs world set-goal --goal "..."
hs world search --query "deploy"
hs world archive-stats      # count by entry type
hs world timeline --limit 10
hs world resolve --entry-id <id> --outcome success
hs world trace --entry-id <id>

---

The Workbench

Alongside the world, Hermespace provides a desk for the current turn — FOA, dual decode, skills+MEMORY fabric, neural FOA, and an autonomy grid.

Feature	What it does
Functional Access Workspace	Harness global workspace — hold/summon concepts, silent reasoning, FOA≤4, hub≤25
Focus of Attention	≤4 items, single active goal per turn
Dual Decode	Human gets a short report (line 1 = next action); the model gets one lean inject. Never dump raw inject into chat.
Cube heart (optional)	Soft cable to HermesCube — skip cube_beat on pre_llm when memory.provider=hermescube (MemoryManager already prefetched); pulse_charge / sync_world_beliefs still recharge World; standalone warehouse when Cube absent
Insight (optional)	Soft cable — perceive_card only (≤400), skip on high/protect; usable/lever write-back on desk.meta only; never required
Skills + Memory Fabric	Ranks Hermes skills per goal; injects MEMORY.md / USER.md excerpts
Neural FOA	HERMESPACE_NEURAL_BACKEND=auto — Ollama embeddings when live, hash fallback
Autonomy Grid	Missions, lenses, dream, self-talk, skillbench, title/tree, access gates. Ground-up design.
Pulse Runtime	Desk-aware pocket rhythm — idle tick, dream, world evolve, mission pulse
Desktop View	Sidebar + full page plugin for Hermes Desktop, with Tailscale viewport and socket API

Quick start

hs install                  # Space plugin+skill; offers Cube + Insight
# or: hermes plugins install PabloTheThinker/hermespace --enable
hs ops doctor               # FAIL if Space is broken; WARN if Cube/Insight missing
hs bench week1              # C1 T1 L1 M1 offline fixtures; Q1 NOT RUN without a judge

For a development checkout:

git clone https://github.com/PabloTheThinker/hermespace.git
cd hermespace
./scripts/install_hermes.sh
python scripts/verify_hermes_integration.py

from hermespace import Workbench

wb = Workbench(agent_id="my-agent", session_id="main")
wb.enter()
r = wb.receive_order("Ship the patch", goal="Ship", decision="A — ship", plan=["implement","test"], say="Shipping.", force=True)
print(r["user_reply"])       # → human
ctx = r["model_context"]     # → model (includes world context)

Hermes plugin hooks

Hook	What happens
on_session_start	Initialize session desk/hub and stage first-turn context
pre_llm_call	Select bounded Access Workspace context (user-message only)
post_llm_call	Observe successful native response and update workbench
pre_tool_call	Observe upcoming tool (name only; fail-open, never deny)
post_tool_call	Record bounded tool-name/count telemetry (no payloads)
on_skill_lifecycle	Park skill:{name}:{event} — no skill body
kanban_task_claimed / kanban_task_completed	Park kanban id so the hub moves
pre_verify	Observe a verify gate (fail-open)
on_session_end	Lightweight end-of-turn receipt (Hermes v0.20 semantics)
on_session_finalize	Harvest ≤10s fail-open, world leave, idle maintenance
subagent_start/stop	Track specialist lifecycle for runtime observability

---

How it works

User message
    │
    ▼
 GATE  ── trivial? skip ── material? continue
    │
    ▼
 ENCODE → DESK → PLAN     FOA, load, decision, steps
    │
    ▼
 DECODE                   report (human) + context (model)
    │
    ▼
 BROADCAST                one lean inject (desk FOA; no world dump)
    │
    ▼
 ACT                      Hermes tools / code / ship
    │
    ▼
 SEAL                     episodic + semantic promote + world evolve

State (local, never committed):

$HERMESPACE_HOME/memory/hermespace/
  ACTIVE.md              live desk
  sessions/              per-session ACTIVE desks (hashed Hermes session IDs)
  access/                per-session AccessHub state
  runtime/               bounded hook telemetry (names/counts, no payloads)
  hermespace.db          turn database
  journal/               human-readable logs
  workbenches/           per-agent workbench JSON
  worlds/                world archive (JSONL) + cache (JSON)

---

CLI

Command	Purpose
hs world show|enter|leave|evolve|search|archive-stats	Persistent world
hs base connect|turn|roles|metrics|lens|audit|chain|harvest	Access Engine (open-source GWT for Hermes)
hs access hold|report|broadcast|lens|swap|audit|reflect|harvest|view	True Access Workspace environment
hs cube status|ensure|beat|pulse|seal|inject	Cube heart/center (standalone-safe)
hs turn	Full INPUT → OUTPUT turn
hs workbench enter|order|idle|park|status	Session workbench
hs fabric / hs skills	MEMORY + ranked skills
hs grid mission-add|lens-set|think|dream|skill-register	Autonomy grid
hs pulse tick|daemon|status	Pulse runtime
hs neural caps|eval	Neural FOA
hs view --serve --port 8764	Desktop viewport
./scripts/smoke_test.sh	Health (9/9)

---

Docs

Doc	Contents
PURPOSE.md	North star — true external Access Workspace + Cube night path
LAYOUT.md	Codespace / folder map
docs/assessment/28-hermes-agent-jspace-assessment.md	Hermes Agent updates → OEW plan
docs/access/thesis-oew.md	Obligatory External Workspace thesis
docs/access/29-baars-changeux-anthropic.md	Baars · Changeux/Dehaene · Anthropic research bridge
docs/access/30-day-to-day-higher-order.md	Day-to-day higher-order Hermes usage
docs/access/31-anthropic-x-video-deep-dive.md	Anthropic X video — how J-space is operated
docs/access/32-hermespace-access-engine.md	Hermespace Access Engine architecture
docs/ops/35-production-operations.md	Hermes v0.20 lifecycle, install, health, recovery
docs/access/27-environment.md	Environment API
ABOUT.md	Philosophy, design principles, author
docs/architecture/CODEMAP.md	Where to edit (layer map)
docs/architecture/HERMESCUBE.md	Cube heart/center contract
docs/architecture/INSIGHT.md	Insight perceive-card cable
INTEGRATION.md	Python · CLI · plugin · workbench doors
skills/hermespace/SKILL.md	Agent skill (load in Hermes)
docs/integration/FOR_HERMES.md	Maintainer / dogfood brief
CONTRIBUTING.md	How to contribute
WORKFLOW.md	GATE → SEAL stages
SECURITY.md	What never ships in git
docs/roadmap/phases-oew.md	OEW phases A–D
docs/README.md	Full docs folder index
spec/DESK.md	Desk schema
spec/PROTOCOL.md	Protocol spec

---

Repository Layout

LAYOUT.md                codespace map (start here for structure)
src/hermespace/          runtime package
  access/                ★ external Access Workspace (hub · env · OEW protocol)
  grid/                  autonomy grid
hermes_plugin/           Hermes session / pre_llm / end hooks
skills/hermespace/       public Hermes agent skill
docs/                    access · assessment · architecture · ops · research
scripts/                 CLI, install, smoke test, security audit
tests/ · experiments/ · desktop_plugin/ · spec/

Full map: LAYOUT.md.

---

Verify

PYTHONPATH=src python3 -m unittest discover -s tests -v
./scripts/smoke_test.sh              # expect pass=9 fail=0
./scripts/security_audit.sh          # before any public push

---

About

Built by PabloTheThinker. Hermespace is an independent companion for Hermes Agent, not an official Nous Research product.

ABOUT.md — philosophy, design principles, related projects.

---

License

MIT