# rtrvr data: train browser agents on knowledge work

> Canonical page: https://rtrvr.ai/data

People use rtrvr for consequential knowledge work across live sites: research, applications, operations, documents, and outreach. We turn 7M+ production runs into trajectories, environments, evals, and verifiers for teams building agents that can finish the work—not just click through it.

## Production proof

- **7M+** live-web runs
- **35K+** users served
- **81.4%** Halluminate Web Bench, ranked #1 ([source](https://rtrvr.ai/blog/web-bench-results))
- **3** complete runs to inspect

## Every run can improve the next

- **Smart DOM:** The page state keeps every interactive element, on screen or off, addressable with a stable ID.
- **Code plan:** The instruction, planner reasoning, generated program, API calls, and resolved arguments stay together.
- **Act + recover:** Typed results preserve errors, retries, human rescue, and the path that finally worked.
- **Verified outcome:** Fresh page-state checks show whether the result—not merely the last action—was complete.

- success → skill
- failure → test
- human rescue → repair
- eval → better agent

## Capabilities this data can measure

- **Perceive the right control:** Pages can contain thousands of interactive elements, including controls outside the viewport.
- **Carry state across systems:** Long tasks cross tabs, redirects, logins, apps, and human approval turns.
- **Recover without restarting:** A stale element or failed tool call should repair the plan, not end the run.
- **Verify the finished job:** The agent must check the result instead of treating its own last action as proof.

## Production task mix

Measured over a recent one-week production window, deduplicated to distinct tasks and grouped by function.

- **Structured research & extraction, 28.8%:** Pull structured records from listings, dashboards, and directories, usually into a sheet. Typical horizon: 10–100+ actions, often fanned across many tabs.
- **Professional applications, 19%:** Applications through Ashby, Greenhouse, Lever, and Workday, including files, fields, and essay answers. Typical horizon: 10–40 actions on one long form.
- **Document & workspace operations, 12%:** Creating and editing docs and spreadsheets, and moving results between the browser and Google Workspace. Typical horizon: 5–50 actions, frequently as the sink of another task.
- **Prospecting & professional outreach, 11.2%:** Finding people, verifying identity, sending connection requests and messages, and confirming delivery. Typical horizon: 5–20 actions per contact.
- **E-learning & assessments, 8.2%:** Course modules, quizzes, and certification flows on SkillsBuild, SCORM players, and publisher courseware. Typical horizon: Very long, often 50–500+ actions across many pages.
- **Business systems & operations, 7.5%:** Government portals, invoicing systems, CRMs, and internal tools, including checkout flows that pause for approval. Typical horizon: 10–60 actions, often multi-step wizards.
- **Publishing in product editors, 6.9%:** Drafting and posting inside the target app's own editor, including CMS bodies and rich-text fields. Typical horizon: 5–25 actions.
- **Social publishing & engagement, 6.4%:** Posting, replying, reacting, and scheduling across social platforms, usually from a list of targets. Typical horizon: 3–15 actions per target, repeated.

## Complete sample runs

### Job application on Ashby

- Prompt: `apply to this job`
- Site: jobs.ashbyhq.com
- Model: deepseek-v4-flash
- 15 steps, 1 plan script(s), 5 sub-agent decisions
- 7 page states containing 722 interactive elements and 61 links
- What it shows: One planner pass writes the program; twelve browser calls execute without another model round-trip. Every field and file upload resolves to the exact interactive element in the captured page state. A fresh page check verifies that the application was submitted.
- Caveat: This is a clean run with no failed actions. Use the Amazon sample for recovery behavior.
- Raw trajectory: https://rtrvr.ai/trajectory-samples/job_application/workflow.json
- Bundle: https://rtrvr.ai/trajectory-samples/job_application.zip

### LinkedIn identity check and outreach

- Prompt: `go send jerry wu of halluminate a linkedin dm of i am about to send you the latest versions of the trajectories`
- Site: linkedin.com
- Model: deepseek-v4-flash
- 4 steps, 1 plan script(s), 5 sub-agent decisions
- 6 page states containing 9,149 interactive elements and 1,551 links
- What it shows: The model checks the search results and resolves the right identity before it sends anything. The selected control is one of 2,040 interactive elements on the profile page. The final page state shows the sent thread, so stopping is tied to visible evidence.
- Caveat: The agent summary says `incomplete` while the final review says `done`. The bundle keeps that disagreement.
- Raw trajectory: https://rtrvr.ai/trajectory-samples/linkedin_dm/workflow.json
- Bundle: https://rtrvr.ai/trajectory-samples/linkedin_dm.zip

### Amazon purchase with human approval

- Prompt: `go order me small splenda pack on amazon`
- Site: amazon.com
- Model: deepseek-v4-flash
- 9 steps, 3 plan script(s), 3 sub-agent decisions
- 5 page states containing 29,915 interactive elements and 4,165 links
- What it shows: The approval question and the human answer are stored inside the trajectory. Two steps fail, the agent repairs the run, and the order is still confirmed. One Amazon page contains 8,699 interactive elements and 1,137 links.
- Caveat: This bundle was redaction-processed: the delivery name, ZIP and order number appear as `[REDACTED_*]` placeholders inside the trees.
- Raw trajectory: https://rtrvr.ai/trajectory-samples/amazon_transaction/workflow.json
- Bundle: https://rtrvr.ai/trajectory-samples/amazon_transaction.zip

All samples: https://rtrvr.ai/trajectory-samples/rtrvr-trajectories-all.zip

Each bundle contains `workflow.json`, `trees/<id>.json`, `trees/INDEX.json`, `SUMMARY.json`, and a schema README.

## Start with the corpus. Build what is missing.

Synthetic tasks are useful for scale. Production runs reveal the instructions, brittle states, human rescues, and end conditions that only appear when people entrust an agent with real work.

### Train on consequential work

Curate production trajectories around the capability, industry, horizon, or failure mode your model needs next.

- Full page states, plans, tool calls, errors, recoveries, human turns, and outcomes
- Cuts for post-training, imitation, offline RL, or targeted error analysis
- New runs, redaction, and delivery format scoped with your research team

### Turn work into an environment

Recreate the starting state, action space, and terminal condition for online rollouts against real browser work.

- Task families and state setup drawn from observed production work
- Browser, API, network, and code actions inside one training environment
- Rewards and terminal checks anchored to the actual end state

### Measure the misses that matter

Build evals around valuable tasks your model is close to solving but still cannot finish reliably.

- Task taxonomies grounded in real knowledge work
- Short actions through workflows that run for 30–60 minutes
- Verifiers for completion, recovery, safety, and efficiency

## Research

- Aug 2025: [SOTA on Halluminate Web Bench](https://rtrvr.ai/blog/web-bench-results). 81.4% across 323 tasks without screenshots.
- Dec 2025: [DOM intelligence architecture](https://rtrvr.ai/blog/dom-intelligence-architecture). How we build a richer, smaller page representation.
- May 2026: [Replacing tool loops with eval()](https://rtrvr.ai/blog/eval-is-all-you-need-agent-harness). Why the planner writes a program instead of one click at a time.
- Jun 2026: [Code as plan](https://rtrvr.ai/blog/code-as-plan-deepseek-flash-text-only-browser-agent). A text-only agent that executes, repairs, and continues.
- Aug 2026: [The self-improving action layer](https://rtrvr.ai/blog/self-improving-action-layer-roadmap). How production runs become better site knowledge.
- Aug 2026: [Nothing but the cache](https://rtrvr.ai/blog/deepseek-prefix-caching-browser-agents). Byte-stable observations and a 90% lower token bill.

## Make one hard task reliable.

Bring one real task your agent cannot finish. We will scope the corpus, environment, eval, or verifier that moves it forward.

Contact: arjun@rtrvr.ai

## Citation

```bibtex
@misc{rtrvr2026trajectories,
  title        = {Web Agent Trajectories from Production Traffic},
  author       = {{rtrvr.ai}},
  year         = {2026},
  howpublished = {\url{https://rtrvr.ai/data}},
  note         = {Production browser-agent runs captured as enriched accessibility-tree
                  observations with typed actions and verified outcomes}
}
```