Open-source design agent: on-brand stills, decks, diagrams, charts and motion graphics from a chat thread, built on Deep Agents.
JavaScript agentagentsai-agentdeepagentsdesignhyperframeslanggraphmanaged-agentsmotion-graphicspaperslackvideo
# Copy to .env (or use `design-agent config set KEY=value`) and fill in what you use.
# `mda deploy` forwards every key here except LANGSMITH_* as a deployment secret: deploy from an
# .env that holds only what the agent uses (no self-hosting keys). See docs/deploy-mda.md.
# Model (provider:model). Gateway form: langsmith:provider/model
DESIGN_AGENT_MODEL=anthropic:claude-opus-5-5
DESIGN_AGENT_MODEL_EFFORT=high
DESIGN_AGENT_HELPER_MODEL=
DESIGN_AGENT_LOOK_MODEL=
DESIGN_AGENT_TOOL_SELECTOR_MODEL=
DESIGN_AGENT_LEARNING_MODEL= # cheaper model for the post-job review (optional)
DESIGN_AGENT_HELPER_EFFORT_CAP= # e.g. xhigh: caps every helper's effort for cost
DESIGN_AGENT_MODEL_TIMEOUT_SECONDS=600 # one call's stall ceiling
DESIGN_AGENT_MODEL_BASE_URL=
DESIGN_AGENT_MODEL_API_KEY_ENV=
ANTHROPIC_API_KEY=
OPENAI_API_KEY= # images (OpenAI)
GOOGLE_API_KEY= # Gemini models; images too when there is no OpenAI key
LANGSMITH_API_KEY= # MDA deploy/dev, LangSmith sandboxes, tracing, optional gateway
LANGSMITH_GATEWAY_API_KEY=
# Runtime
DESIGN_AGENT_RUNTIME=local # local | mda | self_hosted
DESIGN_AGENT_SANDBOX=auto # auto | langsmith | docker | host
DESIGN_AGENT_SANDBOX_IMAGE=design-agent-sandbox:local
DESIGN_AGENT_SANDBOX_SNAPSHOT=
DESIGN_AGENT_SANDBOX_VCPUS=8
DESIGN_AGENT_SANDBOX_MEM_GIB=16
DESIGN_AGENT_SANDBOX_FS_GIB=16
DESIGN_AGENT_SANDBOX_IDLE_TTL_SECONDS=600
DESIGN_AGENT_SANDBOX_EGRESS_ALLOW=
DESIGN_AGENT_MAX_PARALLEL_BUILDERS=3
DESIGN_AGENT_WORKSPACE_ROOT=workspace
DESIGN_AGENT_JOB_BUDGET_USD= # soft: the agent asks before spending past it
DESIGN_AGENT_TIMEZONE=UTC # date note when the channel does not say whose zone
DESIGN_AGENT_CACHE_TTL=1h # lead prompt-cache lifetime (5m | 1h)
DESIGN_AGENT_HELPER_CACHE_TTL=5m
DESIGN_AGENT_MCP_MAX_TOOLS=8 # MCP tools per call for models without native tool search
# Brand
DESIGN_AGENT_BRAND_PATH=workspace/brand # falls back to examples/brands/example-studio
DESIGN_AGENT_PRIVATE_FONTS_DIR=private-fonts
# Media (optional)
DESIGN_AGENT_IMAGE_MODEL=gpt-image-2.5-flare # or gemini-nano-banana-2.1 to make images with Google
ELEVENLABS_API_KEY=
GROQ_API_KEY=
# Paper (optional)
PAPER_MODE=off # off | local | bridged
PAPER_MCP_URL= # local: http://127.0.0.1:29979/mcp or stdio; bridged: https://.../mcp
PAPER_BRIDGE_TOKEN= # bridged; MDA connection "paper-bridge" in production
PAPER_BRAND_FILE_ID= # read-only
PAPER_DRAFTS_FILE_ID=
PAPER_ALLOW_DELETES=0
PAPER_ALLOW_TOKEN_WRITES=0
PAPER_ALLOW_REORG=0 # moving pages and files around
PAPER_ALLOW_GENERATION=0 # paper-gen: bills the person's Paper plan
PAPER_TEAM_ID= # bridge team check
# Optional MCP
EXA_API_KEY=
# Behavior
DESIGN_AGENT_ENABLE_PROGRESS=1
DESIGN_AGENT_ENABLE_FOOTER=1
DESIGN_AGENT_ENABLE_STEERING=1
DESIGN_AGENT_TRACE_HOST= # LangSmith UI host for the footer's trace link; unset, no link
DESIGN_AGENT_ENABLE_LEARNING=1
DESIGN_AGENT_ENABLE_TRACES=1 # the agent reads every thread's traces in its LangSmith project
DESIGN_AGENT_TEMPLATE_TELEMETRY=1 # traces name the template and version; 0 leaves it off
DESIGN_AGENT_ENABLE_TOOL_CLEARING=0
# MDA Slack appearance
DESIGN_AGENT_SLACK_NAME="Content Studio"
DESIGN_AGENT_SLACK_DESCRIPTION="On-brand design and motion from a chat thread."
DESIGN_AGENT_SLACK_BACKGROUND_COLOR=
DESIGN_AGENT_SLACK_TRIGGER_ON_ALL_MESSAGES=0
DESIGN_AGENT_SLACK_ALLOW_BOT_TRIGGERS=0
# Self-hosted
SLACK_BOT_TOKEN=
SLACK_APP_TOKEN=
SLACK_SIGNING_SECRET=
SLACK_TRANSPORT=socket_mode
DATABASE_URL=
POSTGRES_PASSWORD= # self-hosted compose
DOCKER_GID=0 # self-hosted compose: the Docker socket's group
DESIGN_AGENT_API_TOKENS=
DESIGN_AGENT_API_HOST=127.0.0.1
DESIGN_AGENT_API_PORT=8080
# Tests
DESIGN_AGENT_LIVE_TESTS=0 # opt-in tests that call real providers
DESIGN_AGENT_LIVE_VOICE_ID= # an ElevenLabs voice id for the live voiceover test