convez.ai
MCP ServerDesign & UIIntermediário
InnerAura

HyperWeave

InnerAura/hyperweave

Gerador de artefatos visuais SVG para agentes — uma API, sem deps, compatível com tudo

Ver no GitHub
BomPluginScore · de 0 a 100

Camada visual portátil para agentes de IA que gera artefatos SVG auto-contidos em uma única chamada de API. Sem dependências externas, funciona em qualquer markdown e permite que agentes expliquem visualmente o que fizeram, encontraram e modificaram com componentes reutilizáveis e distribuíveis.

Instalação

$npx convez-ai InnerAura/hyperweave

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
61
Forks
10
Acompanhando
2
Saúde
37%
Último commit
há 2 meses
Versão
v0.4.6
Issues abertas
1
Contribuidores
1

Casos de uso

  • Agentes de IA gerando dashboards e relatórios visuais
  • Badges, cards e gráficos para repositórios e documentação
  • Artefatos visuais reutilizáveis para fluxos de automação
  • Identidade visual consistente em múltiplas plataformas (Slack, email, docs)

Veredito

MCP server robusto que resolve um problema real: agentes precisam comunicar visualmente sem overhead. SVG puro, zero JS, funciona em qualquer contexto markdown — excelente para automação e relatórios programáticos. Ideal para times que querem saída visual consistente em agentes.

README

A tradução para o português deste README ainda está na fila.


The Problem

When an agent needs to produce something visual, it generates React, HTML, or pixels: a status card, a dashboard, a receipt of its work, locked in the environment that made it. Embed it in markdown or pass it to another agent and it breaks or loses context. There's no portable visual primitive for agents.

HyperWeave is that primitive. One API call returns a self-contained SVG with data binding, branding, and machine-readable metadata baked in. No JavaScript, no dependencies, no runtime. It renders in GitHub READMEs, Slack, Notion, docs, your site, email, VS Code, or terminal. Every surface that renders an <img> tag is a HyperWeave surface.

FORMATPortabilityAgent-Readable MetadataVisual FidelityToken EfficiencyZero DependenciesSCORE
SVG✓✓✓~✓4.5
MARKDOWN~~✗✓✓3
HTML✗✓✓✗~2.5

Agentic Artifacts

Every AI coding session produces a receipt: cost, tokens, tool distribution, session rhythm. One install, fully automatic.

hyperweave install-hook

The livery matches the agent that produced the session:

AgentLivery
Claude Codeclaude-code
Codexcodex
—voltage
—cream

Auto-detected from the session transcript. Pin a different default with hyperweave install-hook --genome voltage.


Matrices - Generative Tables

HyperWeave matrices are structured tables rendered as portable SVGs. A single JSON description can produce comparison grids, registries, tiers, benchmark tables, numeric heatmaps, chips, glyphs, bars, and status maps, while carrying a machine-readable payload for agents.

READERPIXELSMOTIONREADS VIA
GitHub READMEgithub✓Yescamo, css animation
VS Code previewvscode✓Yesmarkdown preview
Slack unfurlslack~—image proxy
Gmail bodygmail~—img tag
Agentmcp✗—hw:payload, hwz/1, markdown twin

Inside that SVG, alongside the pixels — two tiers, two jobs:

recreate & modify — the complete table IR

<hw:payload schema="matrix/1" media-type="application/json">
{
  "title": "One artifact. Many readers.",
  "subtitle": "how each consumer ingests the same SVG",
  "columns": [
    { "id": "reader", "label": "READER",    "kind": "text",  "align": "left",   "role": "label" },
    { "id": "mark",   "label": "",          "kind": "glyph", "align": "center", "glyph_tint": "full" },
    { "id": "pixels", "label": "PIXELS",    "kind": "check", "align": "center" },
    { "id": "motion", "label": "MOTION",    "kind": "pill",  "align": "center" },
    { "id": "via",    "label": "READS VIA", "kind": "chip",  "align": "left" }
  ],
  "rows": [
    { "label": "GitHub README", "cells": [{ "glyph": "github" }, { "state": "full" }, { "state": "on" }, { "chips": ["camo", "css animation"] }] }
    <!-- … 4 more rows · lossless -->
  ]
}
</hw:payload>

read at a budget — ≈200 tokens to know what an artifact is

<hw:envelope format="hwz/1" media-type="application/json">
{
  "v": "hwz/1",
  "id": "sha256:0077750046fc2780fa4ba19fcd884fcf39f54632306d8c3d2d97f0b7cbf2df47",
  "k": "matrix",
  "title": "One artifact. Many readers.",
  "intent": "structured comparison: One artifact. Many readers.",
  "state": "active",
  "data": {
    "subvariant": "registry",
    "cols": ["mark", "pixels", "motion", "via"],
    "rows": {
      "GitHub README": "github",
      "VS Code preview": "vscode",
      "Slack unfurl": "slack",
      "Gmail body": "gmail",
      "Agent": "mcp"
    },
    "rows_total": 5
  },
  "frames": [{ "t": "matrix", "l": "One artifact. Many readers." }],
  "prov": { "by": "hyperweave", "ver": "0.4.0a2", "genome": "primer.porcelain", "ts": "2026-06-11T15:51:03.185542+00:00" }
}
</hw:envelope>

The envelope is the lossy digest; only the payload round-trips.

  • The round-trip: extract hw:payload, edit the JSON, POST /v1/compose with it as matrix: byte-identical re-render. The envelope's id is the sha256 of the payload, so an agent verifies "this artifact really is this data" before trusting either.
  • The look is a pointer, not a copy: prov.genome: "primer.porcelain" names the aesthetics; payload + that one string is the entire recreation recipe.
  • Markdown twin: every matrix has a GFM projection. --markdown-out on the CLI, respond:"json" over HTTP, render_target="markdown" over MCP.
# Connectors preset
https://hyperweave.app/v1/matrix/connectors/primer.static?variant=porcelain

# Any table, one URL: base64url MatrixSpec JSON (8 KB cap)
https://hyperweave.app/v1/matrix/custom/primer.static?spec=<base64url>

# CLI, with the markdown twin alongside
hyperweave compose matrix --spec-file table.json -g primer --variant porcelain --markdown-out table.md

Another matrice configuration for visualizing benchmarks:

MODELSWE-bench VerifiedINPUT (per Mtok)OUTPUT (per Mtok)
FRONTIER · CLOSED WEIGHTS
Claude Fable 5anthropic95 %$10$50
Claude Opus 4.8anthropic88.6 %$5$25
GPT-5.5openai82.6 %$5$30
Gemini 3.1 Progemini80.6 %$2$12
OPEN WEIGHTS
DeepSeek V4-Prodeepseek80.6 %$0.44$0.87
Kimi K2.6kimi80.2 %$0.95$4
GLM-5zai77.8 %$1$3.2
Mistral Medium 3.5mistral77.6 %$1.5$7.5

SWE-bench Verified % · USD per Mtok · current flagships, jun 2026 · sources: vals.ai · artificialanalysis.ai · model cards


Diagrams - Topologies & Motion

HyperWeave diagrams are topology graphs (pipelines, fan-outs, hub-and-spoke, DAGs, sequences, state machines) rendered as portable SVGs. Like matrices, each carries the full payload / envelope / markdown projection set, so an agent reads the structure, not the pixels. Direction is carried by motion.

hyperweave compose diagram --spec-file /dev/stdin -g primer --variant porcelain --chrome bare -o one-artifact.svg <<'JSON'
{
  "topology": "fanout",
  "title": "One Artifact, Every Surface",
  "subtitle": "a self-contained SVG renders wherever markdown does",
  "notes": "every surface",
  "glyph_tint": "full",
  "nodes": [
    {
      "label": "HyperWeave SVG",
      "desc": "self-contained · portable",
      "role": "hero",
      "glyph": "hyperweave",
      "style": "card+glyph"
    },
    {
      "label": "GitHub",
      "desc": "README · Issues · PRs",
      "glyph": "github",
      "style": "card+glyph"
    },
    {
      "label": "Obsidian",
      "desc": "vault · daily notes",
      "glyph": "obsidian",
      "style": "card+glyph"
    },
    {
      "label": "Slack",
      "desc": "threads · unfurled",
      "glyph": "slack",
      "style": "card+glyph"
    },
    {
      "label": "Email",
      "desc": "inline · PDF export",
      "glyph": "gmail",
      "style": "card+glyph"
    },
    {
      "label": "Agent Context",
      "desc": "hw:reasoning parsed",
      "glyph": "anthropic",
      "style": "card+glyph"
    }
  ]
}
JSON
hyperweave compose diagram --spec-file /dev/stdin -g primer --variant cream --chrome bare -o frontier-serving.svg <<'JSON'
{
  "topology": "dag",
  "title": "Frontier Serving",
  "subtitle": "one key to three frontier labs, shared KV, telemetry skips the ranks",
  "notes": "frontier serving",
  "glyph_tint": "full",
  "nodes": [
    {
      "id": "req",
      "label": "requests",
      "desc": "clients"
    },
    {
      "id": "router",
      "label": "OpenRouter",
      "desc": "one key",
      "glyph": "openrouter",
      "style": "card+glyph"
    },
    {
      "id": "anthropic",
      "label": "Anthropic",
      "glyph": "anthropic",
      "style": "card+glyph"
    },
    {
      "id": "openai",
      "label": "OpenAI",
      "glyph": "openai",
      "style": "card+glyph"
    },
    {
      "id": "gemini",
      "label": "Gemini",
      "glyph": "gemini",
      "style": "card+glyph"
    },
    {
      "id": "kv",
      "label": "kv-cache",
      "desc": "redis",
      "glyph": "redis",
      "style": "card+glyph"
    },
    {
      "id": "obs",
      "label": "metrics",
      "desc": "grafana",
      "glyph": "grafana",
      "style": "card+glyph"
    }
  ],
  "edges": [
    {
      "source": "req",
      "target": "router"
    },
    {
      "source": "router",
      "target": "anthropic"
    },
    {
      "source": "router",
      "target": "openai"
    },
    {
      "source": "router",
      "target": "gemini"
    },
    {
      "source": "anthropic",
      "target": "kv"
    },
    {
      "source": "openai",
      "target": "kv"
    },
    {
      "source": "gemini",
      "target": "obs"
    },
    {
      "source": "router",
      "target": "obs"
    }
  ]
}
JSON
hyperweave compose diagram --spec-file /dev/stdin -g primer --variant porcelain --chrome bare -o mcp-gateway.svg <<'JSON'
{
  "topology": "pipeline",
  "title": "MCP Gateway",
  "subtitle": "host → gateway → server · request and response as two lanes",
  "notes": "mcp gateway",
  "edge_motion": "dash",
  "glyph_tint": "full",
  "nodes": [
    {
      "id": "host",
      "label": "Claude Code",
      "desc": "MCP host",
      "glyph": "claudecode",
      "style": "card+glyph"
    },
    {
      "id": "gw",
      "label": "MCP gateway",
      "glyph": "mcp",
      "role": "hero",
      "style": "card+glyph"
    },
    {
      "id": "server",
      "label": "hyperweave",
      "desc": "MCP server",
      "glyph": "hyperweave",
      "style": "card+glyph"
    }
  ],
  "edges": [
    {
      "source": "host",
      "target": "gw",
      "direction": "both"
    },
    {
      "source": "gw",
      "target": "server",
      "direction": "both"
    }
  ]
}
JSON
hyperweave compose diagram --spec-file /dev/stdin -g primer --variant noir --chrome bare -o frontier-handoff.svg <<'JSON'
{
  "topology": "pipeline",
  "title": "Frontier Handoff",
  "subtitle": "one task relayed across four labs, the comet is the payload",
  "notes": "frontier handoff",
  "edge_motion": "beam",
  "node_style": "glyph-circle",
  "glyph_tint": "full",
  "nodes": [
    {
      "label": "GPT",
      "glyph": "openai"
    },
    {
      "label": "Claude",
      "glyph": "anthropic",
      "role": "hero"
    },
    {
      "label": "Gemini",
      "glyph": "gemini"
    },
    {
      "label": "Ollama",
      "glyph": "ollama"
    }
  ]
}
JSON

Genomes - Aesthetic DNA

A genome is a portable, machine-readable aesthetic specification. It encodes the complete visual identity (chromatic system, surface material, motion vocabulary, geometric form language) as a set of CSS custom properties that any agent can consume and apply consistently across every artifact type.

Four built-in genomes ship today. Custom genome generation via AI skill files coming soon.

brutalistautomatachromeprimer
AestheticRaw materialCellularMetallicMinimal
Variants22 (8 dark, 14 light)16 tones, any two pair5 named8 (4 dark, 4 light)
MotionAnimated border SMILAnimated cell gridAnimated border SMILAnimated state marks
Dividerseam · sigildissolvebandaura

Every broken <img> URL renders the SMPTE RP 219 test pattern with ERR_NNN matching the HTTP status, instead of a browser broken-image icon.


Install

uv add hyperweave            # CLI + SVG rendering (the base)
uv add 'hyperweave[serve]'   # + HTTP server  (hyperweave serve)
uv add 'hyperweave[mcp]'     # + MCP server   (hyperweave mcp)
uv add 'hyperweave[all]'     # + both servers
# or swap `uv add` for `pip install`

Requires Python 3.12+. The base install is CLI + rendering; the HTTP and MCP servers are optional extras so the core stays lean.


Entry Points

Four interfaces, one pipeline. Every path produces the same artifact through the same compositor.

MCP

{
  "mcpServers": {
    "hyperweave": {
      "command": "hyperweave",
      "args": ["mcp"]
    }
  }
}
# Static badge
hw_compose(type="badge", title="BUILD", value="passing", genome="brutalist")

# Data-driven badge — unified token grammar (gh:owner/repo.metric, pypi:pkg.metric, ...)
hw_compose(type="badge", title="STARS", data="gh:anthropics/claude-code.stars", genome="brutalist")

# Strip with multiple live metrics
hw_compose(type="strip", title="readme-ai",
           data="gh:eli64s/readme-ai.stars,gh:eli64s/readme-ai.forks,pypi:readmeai.version",
           genome="chrome")

# Marquee with mixed text + live tokens
hw_compose(type="marquee-horizontal",
           data="text:NEW RELEASE,gh:anthropics/claude-code.stars,text:DOWNLOAD",
           genome="brutalist")

# Discoverable shortcut for single-metric live badges
hw_live(provider="github", identifier="anthropics/claude-code", metric="stars")

hw_kit(type="readme", genome="brutalist", badges="build:passing")
hw_discover(what="all")

CLI

# Badge
hyperweave compose badge "build" "passing" --genome brutalist

# Strip with metrics
hyperweave compose strip "readme-ai" "STARS:2.9k,FORKS:278" -g brutalist

# Live data through the unified --data token grammar
hyperweave compose badge "STARS" --data 'gh:anthropics/claude-code.stars' -g brutalist

# Marquee with mixed text + live tokens
hyperweave compose marquee-horizontal --data 'text:NEW RELEASE,gh:owner/repo.stars,text:DOWNLOAD' -g brutalist

# Artifact kit
hyperweave kit readme -g brutalist --badges "build:passing,version:v0.2.0" --social "github,discord"

# Profile card (live GitHub data, path-segment identity)
hyperweave compose stats eli64s -g chrome -o stats.svg

# Star history chart
hyperweave compose chart stars eli64s/readme-ai -g brutalist -o chart.svg

# Custom genome from a local JSON file (validated against the profile contract)
hyperweave compose badge "DEPLOY" "live" --genome-file ./my-genome.json
hyperweave validate-genome ./my-genome.json

HTTP API

# URL grammar: /v1/{type}/{title}/{value}/{genome}.{motion}
curl 'https://hyperweave.app/v1/strip/readme-ai/brutalist.static?value=STARS:2.9k,FORKS:278'

# Live data via the unified ?data= grammar (works on badge / strip / marquee)
curl 'https://hyperweave.app/v1/badge/STARS/chrome.static?data=gh:anthropics/claude-code.stars'
curl 'https://hyperweave.app/v1/strip/readme-ai/brutalist.static?data=gh:eli64s/readme-ai.stars,gh:eli64s/readme-ai.forks'
curl 'https://hyperweave.app/v1/marquee/SCROLL/brutalist.static?data=text:NEW%20RELEASE,gh:anthropics/claude-code.stars'

# Chromatic variants (automata: 16 solo tones, pair any two via &pair=...; chrome: horizon/abyssal/lightning/graphite/moth)
curl 'https://hyperweave.app/v1/badge/PYPI/automata.static?variant=teal&pair=violet&data=pypi:hyperweave.version'
curl 'https://hyperweave.app/v1/badge/build/passing/automata.static?size=compact'

# Genome-themed dividers
curl 'https://hyperweave.app/v1/divider/band/chrome.static'
curl 'https://hyperweave.app/v1/divider/seam/brutalist.static'
curl 'https://hyperweave.app/v1/divider/dissolve/automata.static'

# Genome-agnostic dividers
curl 'https://hyperweave.app/a/inneraura/dividers/zeropoint'

# Structured frames: /v1/{matrix|diagram}/{preset}/{genome}.{motion}
# (preset 'custom' takes a base64url spec; diagrams add ?chrome=bare)
curl 'https://hyperweave.app/v1/matrix/connectors/primer.static?variant=porcelain'
curl 'https://hyperweave.app/v1/diagram/pipeline/primer.static?variant=porcelain&chrome=bare'

# POST compose
curl -X POST https://hyperweave.app/v1/compose \
  -H "Content-Type: application/json" \
  -d '{"type":"strip","title":"hyperweave","genome":"brutalist","value":"STARS:2.9k"}'

# Local server
hyperweave serve --port 8000

How It Works

Every artifact is the output of a single composition formula:

ARTIFACT = FRAME × PROFILE × GENOME × SLOTS × MOTION × ENVIRONMENT

Python builds context dicts. Jinja2 builds SVG. YAML defines config. Three layers, no mixing. Zero f-string SVG in Python.

ComposeSpec → engine.py → assembler.py (CSS) → lanes.py (validate) → templates.py (Jinja2) → SVG

Every artifact ships with:

  • Semantic metadata: provenance, reasoning, spatial trace, aesthetic DNA. Machine-readable context so the next agent in the chain knows what it's looking at and why.
  • CSS state machines: data-hw-status, data-hw-state, data-hw-regime drive visual transitions through the Custom Property Bridge. No JavaScript.
  • Pure CSS/SMIL animation: all motion uses compositor-safe properties (transform, opacity, filter). No script tags. Works anywhere SVGs render: GitHub's Camo proxy, email clients, Notion embeds.
  • Accessibility: WCAG AA, prefers-reduced-motion, prefers-color-scheme, forced-colors, ARIA markup. Structural, not decorative.
DimensionCount
Frame types11 (badge, strip, icon, divider, marquee-horizontal, stats, chart, matrix, diagram, receipt, rhythm-strip)
Genomes4 (automata, brutalist, chrome, primer)
Motion configs6 (1 static + 5 border SMIL)
Glyphs189 (183 brand marks + 6 geometric shapes)
Divider variants10: 5 genome-themed (band chrome, seam + sigil brutalist, dissolve automata, aura primer) + 5 genome-agnostic (block, current, takeoff, void, zeropoint) at /a/inneraura/dividers/
Metadata tiers5 (Tier 0 silent → Tier 4 reasoning)
Bundled fonts5 (JetBrains Mono, Orbitron, Chakra Petch, Barlow Condensed, Inter), embedded per artifact, no external font requests

Stack: Pydantic, FastAPI, FastMCP v3, Jinja2, Typer.


Data Connectors

HyperWeave binds live data into any artifact through a unified token grammar (?data=...). Tokens are comma-separated; each token is either a literal (text:, kv:) or a live fetch (<provider>:<identifier>.<metric>).

PrefixSourceIdentifier shapeMetrics
gh / githubGitHubowner/repostars, forks, watchers, contributors, issues, pull_requests, last_push, build, license, language
pypiPyPI + pepy.techpackageversion, license, python_requires, downloads
npmnpmpackageversion, license, downloads
crates / cargocrates.iocrateversion, downloads, recent_downloads, license
hf / huggingfaceHugging Faceorg/modeldownloads, likes, tags, pipeline_tag, library_name, license, gated, last_modified
dockerDocker Hubnamespace/repopull_count, star_count, last_updated
arxivarXivid (e.g. 2310.06825)title, authors, published, updated, categories, summary, journal_ref, doi
scorecardOpenSSF Scorecardowner/reposcore (overall trust), plus per-check: code_review, maintained, vulnerabilities, token_permissions, ...
doraGitHub Actionsowner/repodeploy_frequency, lead_time, change_failure_rate, mttr (30-day window)
textliteral—renders the payload as displayed text
kvliteralKEY=VALUEstatic role-tagged value
  • Caching: live values for 5–10 min; a failed fetch caches 60s and shows — rather than a fabricated zero.
  • Isolation: each provider has its own circuit breaker, so one upstream outage can't trip the others.
  • Escaping: commas inside text: / kv: values escape as \,.

→ Open an issue to request a connector.


Contributing

HyperWeave is early. If you're interested in building genomes, extending frame types, or just seeing what this looks like in your own README, join the Discord.


Ver o original no GitHub

Bom para

Relacionados