Beginner-Friendly Tutorials Aug 11, 2026

How to Use GDevelop AI Agent - First Feature in One Evening 2026

August 2026 gdevelop ai agent Monday - Ask AI Build for me, Auto edit on/off, Discord paste, CapEx four-liner, and first-feature receipt before vibe-coding a Steam bet.

By GamineAI Team

How to Use GDevelop AI Agent - First Feature in One Evening 2026

E T Movie pixel art thumbnail

If you are searching gdevelop ai agent on August 11, 2026, you are usually choosing between two vibes: paste prompts into ChatGPT until a browser demo appears, or open GDevelop and let an agent that knows your project build events, objects, and behaviors inside the editor. Weeks later, Discord still pastes “the Agent shipped the mechanic” without a feature card, a remaining weekly %, or a cold Preview — that is why this refresh keeps G1–G6 and adds an explicit Monday ritual plus Discord paste.

GDevelop’s official June 30, 2026 post puts it bluntly: chatbots made the idea → playable box step easy; they did not remove publishing, touch controls, multiplayer, or long-session context collapse. Why Use GDevelop When You Can Just Use AI? The built-in Agent is the answer that stays inside the engine—visible events you can edit after it finishes. GDevelop AI wiki Mid-August is quieter on Twitter and louder in production: July notes still lack a dated Discord paste, and soft “we vibe-coded the feature” claims travel faster than G1–G6 honesty.

This is a one-evening first-feature path—not a Steam save-path freeze week (that challenge lives here) and not a Kimi coding guide (Kimi beginner path). For 3D asset Agents, see Meshy 3D Agent—different pipeline, same “one evening, one artifact” discipline. Phaser Cursor MCP is a different stack entirely (Phaser Game Agent MCP).

Audience Start here Done when
Beginner Feature card → Auto edit off/on + Discord paste You can play one mechanic after a cold Preview
Working dev G1–G6 + credit worksheet Receipt JSON + remaining weekly % written
Studio lead CapEx four-liner + plan-tier honesty Named owner; no-code seat is not a Steam bet
Creator Monday ritual + 15-second clip Pin in Discord before the next jam night

Time: 90–120 minutes for one mechanic you can playtest (or the full Monday ritual).
Honest limit: AI credits are weekly and plan-gated; results vary; you still own playtests and store disclosure.

Facts checked August 11, 2026: Ask AI / Build for me, Auto edit on/off, Medium/High/Maximum reasoning, cloud restore points, and weekly usage reset Sunday 23:50 UTC remain the documented surface (GDevelop AI wiki; Why Use GDevelop). Confirm your remaining weekly % and plan tier tonight—do not trust a July screenshot of someone else’s Ask AI panel. This week’s reset was Sunday August 9; the next reset is Sunday August 16.

Why this matters now (August 11, 2026)

August keep/hold: The June 30 “engine not chatbot” post is six weeks old. The useful question is no longer “does GDevelop have an Agent?”—it is whether your team can finish one numbered feature without turning Ask AI FOMO into a Steam calendar. Soft “Agent shipped the mechanic” claims travel faster than G1–G6 honesty.

Three clocks still drive the evening this week:

  1. Official framing is current, not folklore. Mobile export, free multiplayer, templates, and project-aware AI still beat pure vibe coding past a coffee-break demo. GDevelop blog
  2. Ask AI docs are concrete. Auto edit on/off, reasoning levels, restore points, and weekly usage resets are documented. Wiki
  3. Tuesday is mid-week on the credit clock. Weekly AI % reset Sunday August 9, 23:50 UTC. If a weekend jam already burned High/Maximum on HUD tweaks, tonight is a Medium-only evening—or a hold until August 16. Search still lacks a gated first-feature playbook with that honesty.

If you already filed a July note without a Discord paste or CapEx four-liner, re-run the Monday ritual before partner decks invent “no-code Agent shipped the demo.”

Open-source demand stays loud: 4ian/GDevelop sits around ~25k GitHub stars—the Agent ships inside that same editor ecosystem. Pair the “write it down” habit with BUILD_RECEIPT evenings — different artifact, same culture.

Monday ritual - August GDevelop AI Agent keep/hold (90–120 minutes)

Use this when a lead says “we already tried Ask AI” but nobody can point at a feature card, a remaining weekly %, or a named playtest owner.

Minute Action Pass if…
0–10 Write three-line feature card + check weekly % vs Aug 16 reset (G1, G5) Numbers + do-not list exist
10–25 Save / duplicate; Auto edit off plan chat (G2) Numbered plan you recognize
25–70 New chat; Auto edit on; Medium build (G3) Stop used if it drifts; no silent half-build
70–95 Cold Preview happy / edge / restart (G4) Three paths pass
95–120 Read events; CapEx four-liner + Discord paste + receipt date 2026-08-11 Steam bet still frozen

Standup checklist (ten minutes when you cannot spare the full ritual): Feature card written? Weekly % noted? Cold Preview? Discord paste?

Company CapEx four-liner (diligence zip cover)

When finance asks “should we buy Gold/Pro seats / bet Steam on Agent features?”:

  1. Scope — one-feature evening only; not a whole-game rewrite (GDevelop AI wiki).
  2. Cost — current plan tier + remaining weekly % confirmed tonight; July credit screenshots are not an SLA (wiki — Cost of AI requests).
  3. Evidence — G1–G6 + Preview clip; events read on your project, not a marketing GIF.
  4. Owner — named design lead who freezes Steam copy until disclosure (G6) is written.

Discord paste (copy)

GDevelop AI Agent — Aug 11 2026 Monday ritual (NOT a Steam bet)
Feature: ________________  (numbers: ______)
Do not change: ________________
Plan chat (Auto edit off): done / hold
Build chat (Auto edit on, Medium): done / Stopped / restored
Cold Preview: happy / edge / restart = pass / fail
Weekly AI % after: ____  (next reset Sun Aug 16 23:50 UTC)
Plan tier: Free / Silver / Gold / Pro
Steam disclosure needed: yes / no
Owner: ________  Receipt date: 2026-08-11

August agent-week cadence (after the first evening)

Once G1–G6 pass on one feature, do not auto-queue an RPG system because “the Agent works.” Use a light loop:

Cadence Action
Next evening Second starter card in a new chat — prove Medium still holds
Mid-week (today) Re-open weekly % — if under 30% before Friday, hold High/Maximum until Aug 16
Phaser night Browser MCP stays on Phaser Game Agent MCP — do not stack
Tool-list night Stack shopping stays on 12 AI game-maker tools — different keyword
Steam week Save-path freeze stays on the 7-day challenge

Keep / hold / rewrite (say it in one sentence)

Call Meaning Use when
Keep Agent is your evening mechanic tool G1–G6 pass; events readable; weekly % leftover
Hold Watch credits / plan tier; do not widen scope Mid-week % already thin; High failed twice; jam week
Rewrite Stay on hand-authored events (or leave GDevelop) You need C++/GDExtension, or Agent keeps wrecking do-not lists

Glossary - words the UI will throw at you

Term Plain meaning
Ask AI Top-right panel where you chat and request builds
Build for me / Agent Mode that edits your project (scenes, objects, events)
Auto edit on Applies changes as it works (default for builds)
Auto edit off Proposes; you confirm Yes / Yes+enable / No
Reasoning level Medium / High / Maximum — deeper think costs more credits
Project saved Cloud restore point before an AI request
Weekly AI % Remaining allowance; resets Sunday 23:50 UTC
Official extension Reviewed pack the Agent can install; community packs are not assumed

What the GDevelop AI Agent actually is

Mode What it does Use when
Ask / chat Answers questions about features using your project context Learning, debugging concepts
Build for me (Agent) Creates/edits scenes, objects, behaviors, variables, events, official extensions/assets Shipping a mechanic tonight
Auto edit on Applies changes as it works (default) You already agreed the plan
Auto edit off Proposes changes; you confirm each one Brainstorming without wrecking the scene

Unlike ChatGPT pasting GDScript-ish guesses into a void, GDevelop AI is specialized for the engine and can read your objects, behaviors, variables, and events. Wiki It knows built-in features and official extensions—not arbitrary community packs unless approved.

What it can physically change in your project

Per docs, when it builds it can: create/delete scenes; add/edit/remove objects and place them; add behaviors; organize groups; read/create/delete variables; inspect/adjust effects; create/modify events; create global or scene-wide elements; find and install suitable official extensions or assets. Wiki — What the AI can build

Well suited for: simple tweaks (“make object 25% bigger”), mechanics (“enemy shoots within 20px”), experiments, and learning by reading the events it writes.
Not magic for: fully interconnected RPG systems in one prompt, community-extension-only pipelines, or “make it fun” with no numbers.

Prerequisites

  • A free GDevelop account (browser editor is enough).
  • A project with one scene and a player object (template is fine).
  • 90–120 minutes without Discord tab-fighting.
  • Willingness to ship one mechanic (patrol enemy, coin score, door unlock)—not an RPG system.
  • Optional: cloud project if you want automatic restore points before each AI request. Wiki

Feature card template (copy)

Feature: ________________________________
Numeric values (speed, range, count): ____
Do not change: __________________________
Done when (playtest): ___________________
Reasoning level tonight: Medium (default)

Beginner path - first feature in one evening

Step 1 - Write a three-line feature card

Example (copy and edit):

Feature: Enemy patrols between markers A and B, chases player within 200 px, returns when out of range.
Do not change: Player controls, camera, existing UI.
Done when: I can die once and restart without broken events.

Vague prompts like “make an RPG” waste credits. Official guidance: be specific with values. Agent blog

Step 2 - Save or duplicate before Auto edit

  • Cloud projects: GDevelop saves restore points before AI requests—know where Project saved restore lives. Wiki
  • Local projects: File → save a dated copy. Automatic restore is not available the same way.

Restoring overwrites the current project state—confirm only when you mean to discard later work. Wiki

Step 3 - Plan with Auto edit off

  1. Click Ask AI (top right).
  2. Turn Auto edit off.
  3. Paste your feature card and ask for a numbered plan only.
  4. Negotiate until the plan names objects, behaviors, and events you recognize.

When Auto edit is off and the AI wants a change, you get Yes / Yes and enable auto-edit / No. Wiki Planning with Auto edit off also saves weekly AI usage.

Step 4 - Build with Auto edit on

  1. Start a new chat for the build (one chat per task).
  2. Turn Auto edit on.
  3. Say: “Build the agreed plan. Do not change player controls or camera.”
  4. Watch step statuses (Exploring / Editing / Searching the asset store / …). Click Stop if it drifts.
  5. Prefer Medium reasoning first; raise only if the Agent fails twice. Wiki

For involved requests the AI lays out a numbered plan with pending / in progress / done statuses—expand a step to see detailed actions. If a step struggles on complex logic, it may try an alternative and keep going; you still own the Stop button. Wiki

Step 5 - Playtest like a stranger

Run Preview and check:

  • Happy path (patrol → chase → return)
  • Edge path (player never enters range)
  • Restart path (scene reload does not duplicate enemies)

If broken, do not start a new mega-prompt. Stop, restore if needed, rephrase one failure.

Step 6 - Read the events (the learning tax)

Open the event sheet the Agent wrote. Rename cryptic objects. Delete dead events. This is how beginners become developers—marketing demos skip this step.

Step 7 - Log credits and call G5

Note remaining weekly % after the build. If you burned High/Maximum on a coin pickup, write that mistake down so jam week does not repeat it.

Three starter feature cards (pick one)

Card Good for Trap to avoid
Patrol + chase 200px First combat AI feel Asking for pathfinding + dialog trees in the same chat
Coin score + UI text Feedback loop learning Letting Agent redesign entire HUD
Door unlock with key Inventory-lite logic Spawning infinite keys on restart

Ship one card tonight. Queue the others for new chats on other evenings.

Prompt patterns that waste fewer credits

Pattern Example Why it helps
Numbers first “Chase within 200 px at speed 120” Less guesswork
Do-not list “Do not change Player controls or Camera” Protects working systems
Acceptance test “Done when restart does not duplicate Enemy1” Gives you a Preview script
One failure rephrase “Only fix double spawn on scene reload” Beats a new mega-prompt
Plan then build Separate chats: plan (off) → build (on) Saves weekly %

Developer path - G1–G6 gates before you call it done

Gate Pass when Fail signal
G1 Scope Feature card has numeric ranges and a “do not change” list “Make it cooler” only
G2 Plan Auto-edit-off chat ends with an agreed numbered plan Build started mid-brainstorm
G3 Build Auto-edit-on chat finishes without silent half-builds (or you Stopped and restored) Scene half-rewritten, no Stop
G4 Playtest Three paths above pass on a cold Preview “Works once if I don’t restart”
G5 Credits You logged remaining weekly % and did not burn High/Maximum on a coin pickup 0% left mid-jam for no reason
G6 Disclosure If Agent-built art/audio/narrative ships to Steam, store truth matches Steam AI disclosure keep/hold Labels contradict Content Survey

G1–G6 deep notes

G1 — If you cannot write numbers, you are not ready for Auto edit on. Use Ask AI chat only to learn the feature, then write the card.

G2 — Expand the Agent’s plan steps. If it invents objects you did not ask for, negotiate before enabling Auto edit.

G3 — Watch for “Searching the asset store” when you only wanted logic—Stop and rephrase “no new assets; events only.”

G4 — Cold Preview after close/reopen beats a warm scene that still has editor ghosts.

G5 — Medium is available to everyone; High needs Gold/Pro; Maximum needs Pro. Start Medium; raise only when stuck. Wiki

G6 — Events you would have written by hand ≠ player-consumed gen art. When labels matter, use the Steam AI disclosure debate keep/hold—not vibes.

Receipt mindset (gdevelop_ai_agent_first_feature_receipt_v1.json):

{
  "project": "your-game",
  "feature": "enemy_patrol_chase_200px",
  "auto_edit_plan_chat": true,
  "auto_edit_build_chat": true,
  "reasoning_level": "medium",
  "playtest_paths": ["happy", "edge", "restart"],
  "weekly_ai_percent_after": 72,
  "steam_disclosure_needed": false,
  "gates": {
    "G1_scope": true,
    "G2_plan": true,
    "G3_build": true,
    "G4_playtest": true,
    "G5_credits": true,
    "G6_disclosure": true
  },
  "reviewed_at": "2026-08-11"
}

When you later promote a Steam demo, pair this with GDevelop save-path help—Agent features do not fix Storage roots.

Credit budget - company / creator diligence

Official wiki notes:

  • Usage shows as a weekly percentage near the message box; resets Sunday 23:50 UTC.
  • Higher reasoning levels cost more; Medium is the default balance.
  • Extra usage via subscription tiers (Silver / Gold / Pro), purchased credits, creator feedback, or promo credit claims. Wiki — Cost of AI requests
  • High demand can slow or fail a response—GDevelop says you are not charged when that happens. Wiki

Company note: treat weekly AI % like a cloud budget line. Do not run Maximum reasoning on UI tweaks during a jam week. On Tuesday August 11, you are already two days into the week that reset August 9—if the weekend burned the pot, hold until August 16.

Creator note: pin your remaining % in Discord after each evening so co-devs do not accidentally drain the week. The Discord paste above is the artifact; a thumbs-up emoji is not.

Privacy note (from GDevelop): the AI is not trained on your individual projects; projects are only accessible while you use the conversation; it has no access to profile/leaderboard analytics; it cannot browse the internet; private data is not used to improve the AI. Wiki — AI privacy

Plan-tier reasoning matrix (do not guess)

Level Who can use it Use tonight if… Hold if…
Medium Everyone First feature, HUD, coin, door, patrol You already failed twice on the same card
High Gold / Pro Medium failed twice on a numbered plan Weekly % under 30% before Friday
Maximum Pro Large multi-object request after High failed Jam week, UI tweak, or no restore point

Confirm the current mapping on the wiki Cost of AI requests before you buy a seat because a July blog said “Maximum is cheap.”

Weekly credit worksheet

Item Fill
Week start %
Plan chat cost (estimate) Auto edit off
Build chat cost (estimate) Auto edit on + Medium
Failed attempts
End %
Next Sunday reset noted? Yes / No

Failure modes - what breaks first

Symptom Likely cause First fix
Scene silently rewritten Auto edit on during brainstorm Restore; plan with Auto edit off
Duplicate enemies after restart No acceptance test on reload One-failure rephrase; G4 again
Credits gone after tiny tweak High/Maximum on Medium work Drop level; new chat next week
Agent installs unwanted assets Prompt too open “Events only; no asset store”
Community extension missing Agent does not know unapproved packs Use official extensions or build without
“It works in editor, fails Preview” Warm scene state Cold Preview; G4
Restore wiped good work Restored too far back Confirm restore dialog; keep dated local copies

Common mistakes

  1. One chat for everything — context mush; start New chat per feature.
  2. Auto edit on while brainstorming — accidental scene rewrites.
  3. “Make my game fun” — no numbers, no acceptance test.
  4. Skipping restore discipline on local projects — no undo story.
  5. Assuming community extensions work — Agent prefers official extensions. Wiki
  6. Shipping without reading events — you cannot debug what you never inspected.
  7. Confusing this with Steam save freezes — different job (7-day challenge).
  8. Never rating answers — 👍/👎 (plus a short “why” on downs) is how GDevelop steers the Agent without training on your projects. Wiki
  9. Merging Meshy/Kimi Agent nights into the same receipt — different tools, different gates.
  10. Skipping Steam labels when Agent-made art ships — use the disclosure playbook.

Agent vs ChatGPT / Kimi / Meshy - keep/hold

Stack Keep when Hold when
GDevelop AI Agent You want editable events inside GDevelop + one-click export You need custom C++/GDExtension pipelines
ChatGPT / Claude chat Engine-agnostic brainstorming You expect it to publish APK/iOS for you
Kimi / coding agents You are writing engine code outside GDevelop You wanted no-code event sheets
Meshy 3D Agent You need a game-ready 3D asset evening You needed 2D event logic in GDevelop

For a prompt-first Android no-code overview, see How to Create a Complete Android Game with One AI Prompt. For coding-model weekends, use the Kimi guide. For asset Agents, use Meshy 3D Agent.

Creator checklist - Discord to Preview

  1. Paste the three-line feature card in Discord so collaborators do not widen scope mid-build.
  2. Announce Auto edit off → plan done → Auto edit on → build.
  3. Post a 15-second Preview clip of happy / edge / restart.
  4. Paste remaining weekly AI % after the evening.
  5. If any Agent-generated art/audio ships later, open the Steam AI disclosure keep/hold before store text freezes.

Event-sheet review worksheet (the learning tax, timed)

After G4, spend ten minutes on the sheet the Agent wrote. This is not optional flavor—it is how a beginner becomes the person who can hotfix on Steam week.

Minute Action Pass if…
0–3 Rename one cryptic object / group Name matches the feature card
3–6 Delete or comment one dead event Preview still passes restart
6–8 Find the numeric value (200 px, speed 120) It matches the card—or you wrote the new number
8–10 Screenshot the sheet into Discord under the paste Co-dev can find the events without Ask AI

If you cannot find the chase-range number, you do not own the feature yet—do not call G5 done.

Chromebook / thin-laptop evening notes

GDevelop’s browser editor is the honest first path for Chromebooks and 8 GB laptops. Two traps still wreck August evenings:

  1. Tab death mid-build. Auto edit on + a flaky tab is a half-written scene. Prefer the desktop app if you have it; if you must stay in-browser, one GDevelop tab only, cloud project on, and Stop is faster than refresh.
  2. Restore vs local download. Cloud restore points exist before AI requests on cloud projects. A “Download as file” zip you made in July is not the same as tonight’s Project saved restore. Wiki

Monday creator ritual (10 minutes)

Use this on Monday August 17 (or any Monday) after tonight’s full ritual—not instead of it.

Minute Action
0–2 Check weekly AI % vs Sunday reset (next: Aug 16, then Aug 23)
2–6 Open last Agent event sheet; rename one cryptic object
6–10 Queue next feature card or mark hold

Search, beginners, and what outsiders should cite

Audience What this URL gives you
Beginners Feature card → Auto edit off/on → Preview without chatbot-only demos
Developers G1–G6 gates, receipt JSON, failure modes, reasoning-level budget
Creators Discord pin + Monday ritual + credit worksheet
Companies Weekly AI % as a budget line; privacy notes from official wiki
Search Primary keyword gdevelop ai agent with FAQ for Ask AI, credits, Steam disclosure

Cite-worthy hooks (safe for newsletters / partner footnotes):

Key takeaways

  1. gdevelop ai agent still matters on August 11, 2026 — run the Monday ritual before partner decks invent a Steam bet.
  2. GDevelop AI Agent builds inside your project—objects, behaviors, events—not a detached chat file.
  3. June 30 framing: chatbots speed prototypes; engines still own shipping and iteration. GDevelop blog
  4. Plan with Auto edit off; build with Auto edit on.
  5. One New chat per feature; click Stop when it drifts.
  6. Start Medium reasoning; raise only after two failures. This week reset August 9; next reset August 16.
  7. Cloud restore points ≠ local projects—save copies yourself.
  8. Weekly AI % resets Sunday 23:50 UTC—budget it; Tuesday mid-week is already spent if the weekend jam ran Maximum.
  9. Read and rename Agent events before you call the feature done.
  10. Steam labels are separate—use the disclosure keep/hold when player-facing AI content ships.
  11. Save-path Steam week is a different URL—do not merge jobs.
  12. A July thumbs-up without a dated August 11 Discord paste does not clear CapEx diligence.
  13. Rate answers with 👍/👎 so the team can steer without training on your projects.
  14. Ship one starter feature card per evening—patrol, coin, or door—not all three.

FAQ

Where is the GDevelop AI Agent?

Open your project and click Ask AI in the top-right corner. Use Build / agent-style requests when you want the project edited; use chat when you only want answers. Wiki

Does the Agent replace learning events?

No. It accelerates scaffolding. You still verify Preview and read the event sheet. That is the dual-audience bar.

How do I avoid burning all weekly credits?

Plan with Auto edit off, keep chats focused, stay on Medium reasoning for small features, and stop runaway builds early. Wiki

Can it use every Extension Store pack?

It knows built-in features and official reviewed extensions. Community extensions may not work unless approved. Wiki

Do I disclose Agent work on Steam?

If the Agent only wrote events you would have written by hand, treat it like tooling. If it generated player-consumed art/audio/narrative that ships, follow Steam Content Survey honesty—see Steam AI disclosure debate.

Is this the same as the GDevelop Steam save-path challenge?

No. That challenge freezes Storage paths before fest upload. This article ships your first Agent feature in one evening.

What reasoning level should I start with?

Medium—available to everyone and balanced for most requests. Raise to High/Maximum only when the request is large or Medium failed twice. Wiki

Cloud or local for a first evening?

Cloud if you want automatic Project saved restore points. Local works if you save dated copies before Auto edit on. Wiki

How do I help improve the Agent?

Rate answers with 👍 or 👎; on downs, add a short note why it was wrong. Community projects are not used to train it. Wiki

Do I need to re-run if I already filed a July note?

Yes — if your only artifact is a July note without Discord paste, remaining weekly %, or G1–G6. Soft “the Agent shipped the mechanic” still invents a Steam bet. Ask AI docs did not make Auto edit, credits, or disclosure optional because the calendar flipped.

Is this the same as Phaser Game Agent MCP?

No. Phaser MCP is a Cursor + browser-game Agent evening. GDevelop Ask AI edits event sheets inside GDevelop. Different editor, different receipt. See Phaser Game Agent MCP.

Related reads

Sources

Conclusion

gdevelop ai agent still matters on August 11, 2026: a spike-bounded evening—feature card, Auto edit off then on, cold Preview, weekly % log, receipt—and “Agent shipped the mechanic” claims that travel faster than G1–G6 honesty. Run the Monday ritual, paste Discord, and file the receipt before the August 16 reset invents a panic Maximum run. Hold Steam bets until owners, credit honesty, and disclosure exist. Bookmark this keep/hold beside GDevelop’s Why Use GDevelop When You Can Just Use AI? and the Ask AI wiki. When Steam week arrives, switch to the save-path freeze.