Agent workflows you can see

Draw the curve. Run it as chat or as an API.

Splinel is an open-source visual workspace to build, deploy, and manage AI agent workflows. Models, tools, memory, and the lock that keeps a path runnable after the demo.

npx splinel GitHub

Four points on the curve

One workspace, four locks.

Connect tools, ground the agent in memory, compose logic, then watch the dollar on every run.

01

Integrate

A catalog of apps and models. 45 connectors on this site, counted from the live list on Integrations.

02

Context

Tables, files, and semantic memory the agent can query before it writes a reply or calls a tool.

03

Build

Canvas, conversation, or code. Describe a flow, review the graph it proposes, then edit by hand.

04

Monitor

Traces, logs, and cost per run. Replay a failed step without guessing which model burned the budget.

On the catalog

OpenAIAnthropicOllamaSlackGmail HubSpotSalesforcePineconeNotionPostgres StripeZendesk

Run 8f21 · Path 014

See the dollar on every block.

Splinel traces each node with latency and spend. Model tokens pass through at provider cost. Compute Splinel hosts is itemized on the Path plan.

How monitoring works

classify
$0.0041
crm.lookup
$0.0000
draft
$0.0186
review.lock
$0.0000
send.mail
$0.0004

Total this run $0.0231 · 1.84s · version 3 locked

“We can see the dollar on every run.”

Anika Shah, Platform Lead, Northwind Logistics

“The path we drew on Tuesday is the API on Thursday.”

Jonah Hale, Founding Engineer, Riverline Health

Lock a path this week.

Open the canvas, wire five nodes, deploy as chat or as an API.