Multimodal
Handle text, image, and document flows as first-class runtime inputs.
Multimodal Agent Runtime
Build multimodal agents in Go.
Animus gives you the runtime primitives missing from raw model SDKs: tools, MCP, sandbox, skills, subagents, and artifact-first pipelines designed for real execution.
CLI, service, or embedded runtimeArtifact-first pipelines and lineageSafety surfaces for tools and sandboxing
runtime preview
agent := animus.New(api.Config{
Model: "claude-sonnet",
Sandbox: sandbox.Policy{Filesystem: "workspace-write"},
})
response, err := agent.Run(ctx, api.Request{
Prompt: "Inspect the repository and build a pipeline summary",
Pipeline: pipeline.Batch(
pipeline.Step("scan"),
pipeline.Step("classify"),
pipeline.Step("timeline"),
),
})input.accepted00.00s
tool.exec.repo_scan00.42s
artifact.persist.summary00.89s
timeline.commit01.14s
Tooling
Run built-in tools, custom tools, and MCP-connected tools with control.
Subagents
Coordinate specialized agent paths without collapsing into prompt spaghetti.
Pipelines
Track artifacts, cache, checkpoints, lineage, and timeline in one model.
Why Animus
Runtime primitives instead of another orchestration facade.
Animus is meant for systems that need control, inspection, and deployment fit. The site should make that obvious in one scroll.
More than a model wrapper
Animus turns model calls into a runtime with governance points, orchestration layers, and visible execution flow.
Go-native deployment shape
Use it in CLIs, services, CI systems, and platform backends without dragging in a heavy runtime stack.
Runtime visibility built in
Events, hooks, checkpoints, and lineage make it easier to inspect what ran, what changed, and why.
Execution Model
Show the runtime, not just the slogan.
The visual language centers on code, event flow, artifacts, and control surfaces. The site should feel like an execution system, not a generic AI landing page.
Present external execution as a governed layer with explicit boundaries, not a magical side effect of prompting.
Highlight coordination and specialization without over-selling a black-box platform.
Emphasize why multimodal work benefits from lineage, cache keys, and resumable execution.
Input
Prompt + images + documents
Runtime
Agent loop, model, middleware, tools
Artifacts
Timeline, lineage, checkpoints, cache
Examples + Docs
Examples act as proof. Docs act as the map.
The site treats examples and docs as first-class product surfaces, not as secondary links hidden in a footer.
Basic Agent
Minimal request-response flow with a clean SDK entry point.
Multimodal Input
Pass image and document inputs into the runtime without bolting on side channels.
Artifact Pipeline
Build pipeline steps with cache, checkpoints, and lineage-aware outputs.
Execution Timeline
Inspect the timeline emitted by a pipeline-backed multimodal run.
Getting Started
Understand where Animus fits and how to approach the runtime in one pass.
CLI
Learn how the CLI exposes runtime behavior for streaming, REPL, and pipelines.
SDK
See how model adapters, tools, and orchestration surfaces fit together in Go.
Pipelines
Artifact-first pipeline execution with timeline, cache, checkpoints, and lineage.
Open Source Credibility
Built around real SDK surfaces, not a mock product shell.
Open source Go SDK
Public-facing content should stay tightly coupled to actual runtime capabilities and examples.
CLI and HTTP integration surface
Public-facing content should stay tightly coupled to actual runtime capabilities and examples.
Examples for multimodal, hooks, tasks, and pipeline flows
Public-facing content should stay tightly coupled to actual runtime capabilities and examples.