Brief 7 — Phase 2: Reply desk (watch accounts → surface posts → draft replies)

For: weebo / runbyagent / systemctl (Gocha Berulava)
Sealed: 2026-09-04 (Asia/Tbilisi) · research only · no inbox / threadbus / WakeParent
Question: Is a tool that watches a list of accounts, surfaces posts within minutes, and drafts a house-style reply a viable product — given claimed unit economics via X read-only API (~$0.005/post)?
Inputs reused: /workspace/brief-6-replies-timing.md, /workspace/brief-6-synthesis.md, /workspace/brief-7/roadmap.md
Tags: [MS] multi-source · [SS] single-source · [OFF] official X · [GAP] unresolved

0. Executive answer (one screen)

1. The job already has products. Closest matches are Pounce, XEngage, XReply, SupaBird Engage, ReachMore, Zexr Reply Radar — watch/filter + AI draft + human send. Full auto-post reply bots (Fireply, Replyal) exist but sit on the wrong side of X Automation Rules / non-API scripting risk.
2. Official X line (April 2026): drafting assistance is fine; unsolicited automated replies are not; AI reply bots that post need prior written X approval; scripting x.com outside the API can mean permanent suspension. ([OFF] automation rules)
3. Sub-1k willingness-to-pay is weak and unproven. Cheap entry exists ($15 one-time, ~$19/mo, pay-per-action). Mid-tier $49–$69/mo is sold to “serious” creators / lead-gen, not to accounts with no revenue. No public conversion data for sub-1k → paid reply-desk. [GAP]
4. Verdict: reply desk is primarily a feature of growth suites; a narrow standalone can work only as a cheap, ToS-safe draft assistant (watch list + draft, human posts) — not as API auto-reply. Roadmap Phase 2 positioning (“we use it first”) is the only honest GTM for now.

---

1. What “reply desk” means (roadmap definition)

From /workspace/brief-7/roadmap.md Phase 2: watch a list of accounts → surface posts within minutes → draft a specific reply from house style + account context; self-use first; unit economics on read-only API (~$0.005/post). Brief 6 context that matters for product design:

---

2. Competitor map (watch + draft class)

2.1 Closest to “reply desk” (standalone or reply-first)

2.2 Growth suites where reply is a feature

2.3 Name-check: requested brands that are not this product

2.4 Competitive synthesis

• Crowded drafting lane at $15–$39 entry.
• Watch-list + ranked opportunities is the scarce UI (Pounce, XEngage, Zexr, SupaBird Engage).
• Full auto-reply vendors advertise survival after a Feb 2026 API reply clamp — but official rules still ban unsolicited auto-replies and non-API site scripting. Treat Fireply/Replyal claims as commercially active, compliance-fragile [MS vendor + OFF rules].
• Hypefury’s exit from X is a structural opening for X-native engagement tools ([OFF pricing FAQ] https://hypefury.com/pricing).

---

3. X rules: reply automation vs drafting (2025–26)

3.1 Official sources (use these, not blog paraphrase)

3.2 What the Automation Rules actually say [OFF]

Drafting assistance (human clicks Reply): Not forbidden. The rules regulate automated activity and apps acting through accounts — not a human pasting an AI draft. Automated mentions/replies — default ban:

“Automating these actions to reach many users on an unsolicited basis is an abuse of the feature, and is not permitted. For example, sending automated replies to posts based on keyword searches alone is not permitted.” Automated replies allowed only if all of:

1. Recipient opted in / clearly indicated intent (e.g. replied to you, DM’d you, or took a stated campaign action) — following you alone is not enough;
2. Easy opt-out, honored promptly;
3. One automated reply/mention per user interaction;
4. Reply is to the user’s original post when campaign-based. AI-powered automated reply bots: Explicit subsection — AI dynamic replies are allowed only with prior written and explicit approval from X (partner / developer portal request). Non-API automation:

“Use non-API-based forms of automation, such as scripting the X website. The use of these techniques may result in the permanent suspension of your account.” Other hard nos: automated likes; bulk aggressive follow/unfollow; spammy duplicative posts; OAuth alone ≠ consent for automated actions through a user’s account.

3.3 Feb 2026 API reply restriction (secondary)

Multiple growth blogs claim self-serve API apps can no longer post replies to strangers unless the author mentioned the app first (Fireply, OpenTweet, Bollard summaries). Treat as [SS/MS industry] until mirrored on docs.x.com changelog with a dated entry. Cross-check: https://fireply.ai/blog/x-api-reply-restriction-2026 · https://opentweet.io/how-to/x-api-rules-for-ai-agents-2026 · https://bollardai.com/resources/x Product implication for runbyagent:

• Safe product shape: read API (or browser read) + draft queue + human publish (Chrome or API for own originals). Matches roadmap “agent drafts, Gocha or agent posts depending on what X allows.”
• Unsafe shape: autopilot stranger replies via browser scripting or unsolicited API replies.

---

4. Unit economics check (~$0.005/post)

---

5. Do sub-1k accounts actually pay?

Skeptical default: mostly no at $49+; maybe yes at ≤$20 / one-time / credits.

---

6. Standalone product vs feature?

---

7. Implications for runbyagent Phase 2

1. Build internal desk on read API + house style (brief-6 craft rules). Success metric = hub reply-back rate / profile visits (roadmap), not reply count.
2. If productized: ship draft + approve, never silent stranger auto-reply; disclose AI; no site scripting.
3. Price experiments (if ever public): credits or ≤$19/mo before any $49 claim; dogfood free.
4. Do not compete on “fully automated replies” — X April 2026 rules + non-API ban make that a suspension product.
5. Differentiate with ICP account maps (brief-1 5k–100k builder list) and specificity scoring — the gap left by generic “Great point!” tools.

---

8. GAPS

1. Official docs.x.com changelog URL for Feb 2026 reply-endpoint restriction (industry blogs only).
2. Live TweetHunter / Typefully / ReachMore pricing pages may differ from 3rd-party tables — re-check at launch.
3. No primary WTP survey for sub-1k.
4. Zexr / Bisonary / XreplyAI exact public pricing incomplete.
5. Traction numbers mostly vendor or single reviews (Fireply PH upvotes, TweetHunter 5.6k, SupaBird “30k posts”).
6. ReplyAgent / SuperReply / EngageAI / Bird / Circleback — no clear X reply-desk product under those names.
7. Whether X API 24h read dedupe applies to user-timeline polling at desk scale (affects $0.005 math).

---

COUNTS