Agentic Visual Orchestration
meLink avo
A visual operating system for AI teams
Design reusable multi-agent workflows on a canvas, validate before you spend tokens, and run with full runtime visibility. Mix cloud and local models, gate risky actions, and inspect every step.
Visual first. Execution-safe.
From canvas to controlled execution
Build orchestrations inside projects, connect agents and gates on a visual editor, then dry-run and execute with a live timeline.
01
Design on the canvas
Drag agents, validators, decision gates, human approvals, and I/O nodes onto a React Flow canvas. Configure provider, model, prompts, tools, and memory per node.
02
Validate before you spend
Dry-run checks structure, missing providers, broken edges, loop risks, and budget estimates. Fix warnings on the canvas before any tokens are used.
03
Run and inspect
Execute mock or live runs with a streaming runtime timeline. See node status, duration, token estimates, approvals, and errors as the workflow moves.
Built for real orchestration work
AVO is a control plane for reusable AI teams: not just a node editor, but scheduling, permissions, validation, and runtime inspection end to end.
A
Agent Library
Save configured agents as reusable templates. Search, drag into new orchestrations, and build coder, reviewer, QA, and research teams without rewriting prompts.
L
Cloud and local models
Connect OpenAI, Anthropic, and local runtimes like Ollama. Set provider concurrency limits and queue GPU-bound calls safely on your own hardware.
R
Local runner
Pair an approved local machine to read and write files, run git commands, and execute sandbox actions through a controlled runner, never unsafe browser OS access.
H
Human approval gates
Pause workflows at approval nodes for risky actions like file deletion, package installs, or git push. Approve, reject, edit instructions, or stop the run.
V
Validators and decisions
Route pass/fail branches with validator nodes. Use AI decision gates to approve or reject JSON output before the workflow continues.
M
MCP and web tools
Attach MCP servers and web search skills to agents. Tool permissions and execution events show up in the runtime timeline.
P
Pipelines and knowledge
Chain orchestrations into pipelines, inject project knowledge with vector RAG, and import or export workflow JSON with secrets redacted by default.
$
Token and cost visibility
See estimated usage per run, per node, and per provider. Dry-run gives a rough range before live execution spends real tokens.
You’ve got this
Model your AI teams visually, validate before you spend, and run with clarity. Talk to us about early access to meLink avo.