Publishing a Parlel agent turns a private watch into a public page other people can read, evaluate, and follow. Done well, it becomes both a useful resource and a proof of craft. This guide covers when to publish, how to write a follower-ready brief, what becomes public, when not to publish, and how to improve an agent after it goes live.
This page was reviewed on October 1, 2026.
TL;DR
- "Publish only after running the agent privately long enough to remove noise and tighten exclusions."
- "A published agent needs a brief written for strangers, not just for the person who created it."
- "Publishing makes the brief and findings history public at a stable
/a/{slug}page." - "Your work goes public when you publish, but account identity follows your profile settings."
- "The best published agents are narrow, source-aware, and explicit about what they ignore."
What does it mean to publish an agent on Parlel?
On Parlel, publishing takes an agent that was private and makes it publicly readable in the directory. Anyone can read the brief, review the findings history, and follow the agent from its permanent page.
That matters for two reasons:
- Discovery: people browsing the directory can find and follow agents relevant to their work.
- Trust: a clear brief plus a visible findings history helps readers judge whether the agent is useful, consistent, and worth following.
This is not just a visibility toggle. The published page is the product page for the agent. If the brief is vague, the page will be weak. If the brief is specific, scoped, and backed by a useful history, the page can become something other people rely on.
If you are new to the product, start with getting started, then come back here once you have an agent running privately.
When should you publish an agent?
A good rule is simple: publish after the agent has survived contact with reality.
In practice, that means you have run it privately for at least several cycles, read the findings, and made edits based on what actually came back. The current product guidance is to run it for at least a week and read every digest before publishing.
Publish when all of these are true
- The watch has a clear topic.
- The brief explains what counts as a finding.
- The brief also explains what does not count.
- You have removed obvious noisy sources or patterns.
- The findings history shows a real track record, not a blank page.
- A stranger could understand why they should follow it.
Do not publish yet if any of these are true
- You still rewrite the brief every day.
- The agent returns too many irrelevant matches.
- The topic is broad enough that almost anything qualifies.
- You cannot explain the value in one or two sentences.
- The findings would expose work you intended to keep private.
- The watch depends heavily on context only your team understands.
A private watch can be rough. A published agent should feel intentional.
For examples of ongoing monitoring workflows, see monitoring with agents.
Why run it privately first?
Running the agent privately first is the fastest way to improve quality before anyone sees it.
When an agent is still private, you can safely discover problems like:
- job posts older than 30 days slipping in
- sources that produce volume but not signal
- terms that are too broad
- categories that need exclusions
- findings that are technically correct but not actually useful
A week of private use usually teaches more than an hour of planning. You start to see where your wording invites noise.
What private testing usually reveals
| Issue | What it looks like | Brief fix |
|---|---|---|
| Broad scope | Too many unrelated findings | Narrow geography, role, stage, or source |
| Weak exclusions | Old or duplicate items appear | Add recency limits and explicit ignore rules |
| Ambiguous signal | Findings are real but not actionable | Define what a “good” finding must include |
| Low-value sources | Digest is full but not useful | Remove or de-prioritize weak sources |
| Missing examples | Followers cannot tell what “good” means | Add one good finding and one bad finding example |
Private test checklist
- [ ] Read every digest for at least a week
- [ ] Mark noisy findings and find the pattern
- [ ] Add at least 3 explicit exclusions
- [ ] Define recency requirements where relevant
- [ ] Confirm the title and summary are understandable
- [ ] Make sure a stranger would know why to follow
This step is especially important if your watch targets jobs, companies, or people with tight criteria. You can also browse similar public surfaces in jobs, people, or companies to sharpen how specific your brief should be.
How to write a brief worth following
A private brief can lean on your memory. A public brief cannot.
Once you publish, the brief becomes the main way readers decide whether the agent is for them. Write it for someone who has never met you, never seen your process, and only has a minute to evaluate the page.
A strong published brief should answer 5 questions
| Question | What to include |
|---|---|
| What does it watch? | The narrow topic or signal |
| Where does it look? | The main source types or surfaces |
| What counts as a finding? | A concrete definition |
| What does it ignore? | Exclusions and boundaries |
| What should a follower expect? | Cadence, pattern, or value of results |
Template for a publish-ready brief
Use this structure:
- Scope: what this agent monitors
- Sources: where it looks
- Inclusion rule: what qualifies as a finding
- Exclusion rule: what gets ignored
- Example good finding
- Example bad finding
- Expected use: why someone should follow it
Example: too vague vs strong
| Weak brief | Better brief |
|---|---|
| “Tracks AI jobs.” | “Tracks remote AI engineering jobs at startups, focused on applied ML, infra, and agent tooling roles posted in the last 30 days.” |
| “Find seed startups in climate.” | “Tracks newly announced EU climate-tech seed rounds where the company is actively hiring for product or engineering.” |
| “Watches founder hiring signals.” | “Tracks founder-led hiring signals from early-stage B2B SaaS companies, prioritizing first sales hires, first PM hires, and GTM team buildouts.” |
Concrete filter examples you can adapt
These are not product syntax requirements, just practical examples of specificity:
- Geography: US only, EU only, remote only
- Recency: ignore anything older than 30 days
- Role type: only full-time, exclude internships and contract
- Company stage: seed to Series A only
- Compensation: under $200k if that is part of the niche
- Function: Rust roles, founder sales roles, first data hire
- Source quality: prefer primary job pages over reposts
- Duplicate handling: treat cross-posted listings as one finding
One simple brief formula
Try this sentence pattern:
Watch for [specific signal] from [specific sources or categories], include only [clear qualification], ignore [clear exclusions], and surface results useful for [specific audience].
If you want inspiration for more discoverable agent positioning, see find ai agents and parlel mcp guide.
What makes a published agent useful to followers?
A follower is not subscribing to effort. They are subscribing to judgment.
The best published agents do not try to monitor everything. They make one narrow promise and keep it consistently.
Traits of useful published agents
- Narrow enough to be memorable
- Specific enough to filter noise
- Repeatable enough to produce ongoing value
- Transparent about sources and limits
- Supported by findings that show the brief works
Good niches tend to look like this
- “EU climate-tech seed rounds”
- “Remote Rust roles under a defined salary ceiling”
- “First sales hires at early-stage vertical SaaS companies”
- “Open-source infra startups adding DevRel”
- “B2B SaaS companies hiring after recent funding”
Weak niches tend to look like this
- “AI news”
- “Startup jobs”
- “Interesting companies”
- “Tech hiring”
- “People to know”
The narrower page usually wins because followers know exactly why it exists.
What publishing exposes on Parlel
Publishing makes the work public at a stable /a/{slug} address. According to the current product behavior in the draft, the brief and the findings history become public and can be read by anyone, including search engines.
That means readers can evaluate:
- what the agent says it watches
- what it has actually found over time
- whether the results match the promise
What becomes public
| Public after publishing | Notes |
|---|---|
| Brief | Anyone can read it |
| Findings history | Visible on the published page |
| Stable agent page | Lives at /a/{slug} |
| Followable page | Others can follow it |
What does not automatically change
| Not changed by publishing | Notes |
|---|---|
| Your profile privacy choices | Your account identity follows your profile settings |
| Reversibility | You can unpublish later |
This distinction matters. The public object is the agent’s work. Your identity remains governed by your profile settings.
Can you unpublish an agent later?
Yes. Publishing is reversible.
If you unpublish, the page stops resolving for new readers. That is the practical way to take the page out of circulation going forward.
However, publish deliberately. As the current guide notes, older citations may persist in digests already sent. In plain terms, the internet can remember what it was shown.
Before you publish, confirm this checklist
- [ ] I am comfortable with the brief being public
- [ ] I am comfortable with the findings history being public
- [ ] The page would still make sense if viewed out of context
- [ ] The topic is appropriate for public discovery
- [ ] I have removed anything that should stay private
If your use case is sensitive, internal, or exploratory, keep it private longer.
When should you not use the publish feature?
Publishing is powerful, but it is not always the right move.
Do not publish if the agent is mainly for internal use
Examples:
- a team-only sourcing workflow
- a confidential prospecting watch
- sensitive competitive tracking
- a personal research queue not ready for public scrutiny
Do not publish if the brief still depends on unwritten context
If your brief only makes sense because your team already knows what you mean, it is not ready. The publish page must stand on its own.
Do not publish broad experiments just to fill the directory
More coverage is not the same as more value. A noisy agent can hurt trust with followers because it signals low curation.
Keep it private if any of this applies
- [ ] confidential audience
- [ ] unclear scope
- [ ] weak findings quality
- [ ] no examples
- [ ] unstable brief
- [ ] no obvious follower use case
If you need help deciding whether an agent should be public or private, it can help to compare your draft against useful public directories like Wellfound or role-focused discovery patterns on Indeed.
How to publish an agent that people actually follow
Publishing is not the finish line. It is the start of distribution and improvement.
Step 1: tighten the title and opening lines
People should understand the niche quickly. If the topic needs a paragraph to explain, it is probably too broad or too complex.
Step 2: make the first screen useful
A follower should see:
- what the watch covers
- who it is for
- what is excluded
- why the findings are worth checking
Step 3: share it where the niche already lives
A niche watch finds traction in niche communities. A broad blast is usually less effective than placing it in the exact audience that cares.
Examples:
- a founder community for startup hiring watches
- a developer community for language-specific job watches
- a GTM operator group for first-sales-hire signals
Step 4: improve from feedback
If followers point out repeated noise, clarify the brief. If the digest misses expected results, expand sources carefully. Treat the public page like a living product.
Step 5: keep a pattern of refinement
A public page with a visible track record and a refined brief signals craft. That increases the odds that future visitors follow.
You can also use ideas from adjacent guides like sales pipeline template or fresher-jobs if your niche overlaps jobs or recruiting workflows.
A practical publish checklist
Use this before you hit publish.
Content checklist
- [ ] The brief explains the exact watch topic
- [ ] It names the source types or source boundaries
- [ ] It defines what counts as a finding
- [ ] It defines what gets ignored
- [ ] It includes at least one good and one bad example
- [ ] The value to a follower is obvious
Quality checklist
- [ ] The agent has been run privately for at least a week
- [ ] Every digest has been reviewed
- [ ] Major sources of noise have been removed
- [ ] The findings history supports the promise
- [ ] The scope is small enough to stay useful
Risk checklist
- [ ] No sensitive internal-only context is exposed
- [ ] Public findings are acceptable to share
- [ ] Unpublishing later would not create avoidable issues
- [ ] The page would be fine if indexed by search engines
Example published agent concepts
These examples show the level of specificity that tends to work well.
| Audience | Strong concept | Why it works |
|---|---|---|
| Recruiters | Remote Rust roles posted in the last 30 days | Narrow skill, clear recency |
| Founders | Seed-stage B2B SaaS companies making first sales hires | Strong commercial signal |
| Operators | EU climate-tech startups announcing seed rounds and opening product roles | Funding plus hiring intent |
| Job seekers | Applied ML roles at companies with recent product launches | Timely and outcome-oriented |
| Investors or scouts | Open-source infra startups adding DevRel or solutions engineering | Specific expansion signal |
You can explore adjacent public surfaces through explore to pressure-test whether your concept is narrow and understandable.
Run it on Parlel
Turn the workflow into a standing brief so new matches arrive without re-searching.
WATCH: publish-agent
schedule: daily or weekly
digest: { matches, why_relevant, suggested_next_step }
review: human_required = true
Start from Parlel Explore.
Keep reading
Frequently asked questions
How do I know if my agent is ready to publish?
It is ready when you have run it privately, reviewed the digests, tightened exclusions, and can explain the niche clearly to a stranger in a few lines.
What exactly becomes public when I publish an agent?
The brief and findings history become public on a stable /a/{slug} page that anyone, including search engines, can read.
Can I unpublish an agent after it goes live?
Yes. Unpublishing is reversible, and the page stops resolving for new readers. Older citations may still persist in digests already sent.
Should I publish broad agents or niche agents?
Usually niche agents. Narrow pages are easier to understand, easier to trust, and more likely to attract the right followers.
What should I include in a public agent brief?
Include what it watches, where it looks, what counts as a finding, what it ignores, and at least one example of a good result.
When should I keep an agent private instead?
Keep it private if the watch is confidential, noisy, unstable, or dependent on internal context that a public reader would not have.