STOP WASTING PAID TOKENS. START POOLING ACCOUNTS TODAY. [ GET YOUR VAULT ]

SpiderGate Studio

The Studio is the redesigned workspace inside your SpiderGate dashboard. It's where you chat with any model, run one prompt across several models at once, generate images, video and audio, and pull your own data straight into the conversation — all in one place.

Open it at app.spideriq.ai/dashboard/gate → Studio.

What you can do

  • Chat with any model, with your real name and avatar on the conversation, formatted Markdown replies, and an optional strict JSON mode.

  • Fork & Merge — send one prompt to two or three models side by side, or merge their answers into a single response.

  • Projects — group related sessions and everything you generate into one workspace.

  • Saved prompts — name a setup once and reload it in one click, or call it from an agent by name.

  • Trace — see exactly which model answered each turn, with tokens, latency and cost.

  • @-mention your own data — bring your CRM and lead records into the chat as context.

  • Media generation — create images, video and audio using your own provider key.

  • Reference images — pick a reference from a visual grid and see it inline in the conversation.

  • Past generations — every project keeps a gallery of what you made, and one click reuses the recipe.

  • Per-model settings — every model shows exactly the controls it supports, nothing more.

Chat

The chat is the heart of the Studio.

  • Real identity. Your name and avatar appear on your messages; each reply is labelled with the model that produced it — no more guessing who said what.

  • Markdown replies. Responses render as formatted text — headings, lists, tables and code blocks — instead of a raw wall of text.

  • JSON mode. Toggle a per-conversation JSON option when you need structured output you can drop straight into code.

  • History. Every conversation is saved. Reopen a past session, keep going where you left off, and find older chats from the history list.

  • Rich model picker. A searchable picker lets you find a model by name or task alias (like spideriq/coding or spideriq/fast) and see what each one is for before you pick it.

Fork & Merge

Sometimes one model isn't enough.

  • Fork runs the same prompt across two or three models at once, side by side, so you can compare answers in a single view.

  • Merge takes those two or three answers and combines them into one consolidated response.

Add models to a fork with the same rich picker used in chat, and every fork or merge session is saved to your history just like a normal conversation.

Projects

Projects keep your work organised. Group the chats, forks and generated media that belong to one piece of work into a single project, and jump back into that context whenever you return.

Saved prompts

A prompt you got right is worth keeping. Save it once and it becomes a reusable, named bundle you can reload in a click — and that your agents can call by name.

A saved prompt bundles four things together, not just the text:

  • the system prompt (the instructions),

  • the model it was written for,

  • the settings (temperature, size, and whatever else that model exposes),

  • any reference images attached to it.

Saving one. Get the composer the way you want it, then open Prompts → Save current setup (or type /save-prompt). Give it a name and an optional description. The model and settings are captured from the composer as-is; you can attach reference images to the prompt from the same picker used elsewhere in the Studio.

Using one. The Prompts gallery lists everything saved in the current project, with its description, who made it, and when it was last used. Pick one and it loads straight back into the composer — instructions, model, settings and references together. Type /prompts to open the gallery from the keyboard, and search by name or description when the list gets long.

Saved prompts live inside a project, so each project keeps its own set. Names are unique within a project — saving a second prompt under an existing name is rejected rather than silently overwriting the first.

Calling a saved prompt from an agent

This is the part that makes a saved prompt more than a UI convenience. Anything holding your API key — an agent, a script, our CLI or MCP tools — can reference a saved prompt instead of repeating its contents, and SpiderGate expands it server-side. Put one of two forms in the prompt field of a chat or media-generation request:

  • prompt:<id> — the stable id shown on the prompt (it looks like prompt_…). It resolves anywhere in your workspace, no project needed. This is the form to prefer when something has to keep working unattended.

  • prompt.<handle> — the human-friendly form, resolved within a project, so the request has to say which project. The handle is exactly the string the save dialog shows you ("Agents reference it as …") — your prompt's name lowercased with punctuation and spaces turned into underscores, e.g. Hero Food Shot → prompt.hero_food_shot. The full name works too.

A reference that doesn't resolve is now an error, not a shrug. If the handle doesn't match anything in the project you get a 404; if two prompts share a handle you get a 409 naming both prompt:<id>s so you can pick one — SpiderGate never guesses. Previously a mistyped reference was quietly treated as ordinary prompt text, so the generation ran without your saved model, settings or reference images and nothing said so.

Tip: anything the request sets itself wins. The saved prompt only fills in the blanks — send your own model or a different temperature alongside a prompt reference and your values are kept. That makes a saved prompt a sensible default you can override per call, not a straitjacket.

Referencing a prompt updates its last used timestamp, so the gallery surfaces what you actually rely on.

Agents can also manage prompts directly — create, list, get, search and delete — through our CLI (spideriq gate prompt …) and the matching MCP tools, so a prompt library can be built and maintained without opening the dashboard.

Trace

Open the Trace tab to inspect any turn: which model handled it, how many tokens went in and out, how long it took, and what it cost. It's the same observability you get from the Traces dashboard, right next to the conversation.

Bring your own data (@-mentions)

Type @ in the composer to pull your own records into the chat as context — businesses, contacts, emails, companies and saved pins from your CRM and lead data. Ask questions about your leads, summarise a contact list, or draft outreach grounded in your real records without copying and pasting.

When a model returns structured results — a list of leads, a table, a set of stats — the Studio renders them as rich cards instead of plain text, so the answer is easy to read and act on.

Media generation (bring your own key)

The Studio can generate images, video and audio. In this first release, media runs on your own provider key — add a key for fal, OpenAI, Google or x.ai in your dashboard, and generations are billed directly on that provider's usage. Everything you create appears inline in the chat and in your project gallery.

Reference images

Some models don't just take words — they take a picture to work from. When the model you've selected accepts one, an Add to prompt button appears in the composer (or type /reference). It opens a visual grid of your media: what's already in this project, your wider media library, and an upload tile for something new.

Pick an image and it's shown inline in the conversation, right where you're working, rather than reduced to a filename chip. After a generation runs, the reference stays visible next to the result, with Use as next reference to carry it into the following prompt.

Note: if the model you've picked can't take a reference image, the button is still there but disabled, and it tells you why. That's deliberate — a missing button looks like a bug, whereas a disabled one with a reason tells you to switch models. Whether a model accepts a reference is decided by the inputs that model actually declares, not by its category label, so trust the button over the wording of the model's name.

Your past generations

Open the Images, Video or Audio canvas in a project and you land on a grid of everything you've already made there, newest first — not an empty box.

  • Reuse on any tile loads that generation's whole recipe back into the composer: the exact prompt, the model, the aspect ratio and the panel settings. Change one word and run it again.

  • Use as reference takes the generation itself and attaches it as the reference image for the next run, so you can iterate on something you made.

  • Remove takes a tile out of the gallery.

The grid is per project and per media type, so the Images canvas shows that project's images. A project with nothing in it yet shows the usual "describe what you want to create" prompt.

Per-model settings

Different models expose different controls. The Studio reads each model's own settings schema and shows you exactly the tunables that model supports — temperature, size, steps, voice, and so on — so you're never shown a setting that doesn't apply.

Where to go next

  • Playground — the quick prompt-and-copy view.

  • Task Aliases — the spideriq/* aliases you'll see in the model picker.

  • Traces — full per-request observability.

  • Agent Keys — manage the keys the Studio uses.