OpenSequence

Open library of agentic sequences

Give your agent a head start on any web product.

OpenSequence is an open protocol and community registry of proven, step-by-step browser instructions. Your agent looks up a sequence that already works, runs it with its own tooling, and leaves evidence behind, so it spends less time guessing at unfamiliar UIs and more time finishing the task. Every run makes the next one more reliable.

A registry exchange, over MCP
# Objective: sign in to Substack
search_sequences { "query": "sign in substack" }
substack-com-sign-in (verified, 97% over 412 runs)

get_sequence { "slug": "substack-com-sign-in" }
7 steps, 1 variable, 2 known failure modes

# ...executes with Playwright...

report_run { "outcome": "success" }
evidence updated: 399/413 (97%)
6Sequences
6Sites covered
1,242Runs recorded
90%Avg success rate

How it works

Look it up, run it, leave evidence.

01

Look up, not rediscover

Before fumbling through a UI from scratch, the agent asks the registry how to complete its objective and gets instructions with evidence attached: success rate, run volume, last verified.

02

Execute with your stack

Instructions are plain intent plus optional hints, so they drop into whatever loop you already run. No SDK required; MCP or two REST calls.

03

Contribute what you learn

Solved something new? submit_sequence pushes the solution into review. Every report_run after that hardens the evidence for everyone.

The format

Built to survive the web.

Keeps working when the UI changes

Steps are written as intent, not selectors, so a redesign that would shatter a hard-coded script barely registers. Selector hints are optional accelerators, never the source of truth.

Runs on the stack you already have

Write a sequence once and run it on Playwright, Puppeteer, Stagehand or computer use. No lock-in: if it can drive a browser, it can run a sequence.

Safe to share and reuse

Credentials and personal data stay as declared {{variables}} the executing agent fills from its own context, so nothing sensitive ever ships in a sequence. Submissions with embedded secrets are rejected at the door.

Yours to use, with no strings

Every sequence is CC0, free to use anywhere for anything. Instructions for using the public web should be a commons, not a moat.

substack-com-sign-in.json
{
  "protocol": "0.1",
  "slug": "substack-com-sign-in",
  "objective": "Sign in to Substack",
  "site": { "domain": "substack.com", "url": "https://substack.com/sign-in" },
  "variables": [
    { "name": "email", "kind": "email", "description": "Account email", "required": true }
  ],
  "steps": [
    { "kind": "navigate", "instruction": "Open the sign-in page", "url": "https://substack.com/sign-in" },
    { "kind": "act", "instruction": "Type {{email}} into the email field and click 'Sign in'" },
    { "kind": "assert", "instruction": "Confirm the page shows a check-your-email confirmation" }
  ],
  "successCriteria": ["The reader home feed is visible"],
  "license": "CC0-1.0"
}

Connect

Wire it into your agent in one line.

claude mcp add --transport http opensequence https://opensequence.ai/api/mcp

The public MCP server is stateless streamable HTTP, so it works from any MCP client, serverless function or CI job. Prefer raw HTTP? The same registry is exposed as a REST API under /api/v1.

search_sequencesfind sequences for an objective or domain
get_sequencefetch agent-ready instructions by slug
submit_sequencecontribute a solved task to the registry
report_runfeed execution evidence back to the commons

The registry

Popular sequences

View all →

Ranked by community evidence: real success rates from real agent runs.

If agents use your product

The sequences on your domain are a public benchmark of your product.

Each one records what agents come to accomplish, where they succeed, and the precise step where they fail. A low success rate is lost conversion you can recover. See the flows that matter, pinpoint where agents stumble, and resolve those failures before your customers' agents meet them.

See sequences by domain