Repository navigation
Releases
2.5.2 — README for people, the Supports list
Changed
- Documentation: the README is one screen (install, try it, the tools by group, what it supports, the docs); the
reference moved todocs/—docs/CONFIGURATION.md,docs/TOOLS.md,docs/PRICES.md(an index of the measured
quotes),docs/TRADEMARKS.md, the development commands toCONTRIBUTING.md;AGENTS.mdholds the commands, the
places and the rules for coding agents (CLAUDE.mdpoints at it). The README's "Supports" list names the version it
was checked for, andnpm run check:supports— run by CI and by the release workflow before staging — fails while
that version differs frompackage.json's. Nothing in the published server changes.
Development
- The repository's lockfile:
npm audit fix—qs6.17.1 (was 6.16.0,
GHSA-6pq8-jhm2-2vcw) and five moderate advisories;npm audit
on the repository reports none.package-lock.jsonis not part of the published package and the dependency ranges
inpackage.jsonare unchanged, so an install resolves its own tree as before.
2.5.1 — structure lint, Prettier, dependency updates
Changed
- Development only:
npm run lintchecks the structure of the code with a shared preset (ESLint with Babel's
TypeScript parser — TypeScript 7 has no JavaScript API for typescript-eslint —, jscpd for repeated blocks, a ratchet
that keeps the recorded baselines from growing); CI runs it. Nothing in the published package changes. - Internal: the usage, raw-call, file-answer and training tools each hand one step to a function of its own (the usage
query, the upload fields, the failed file answer, the training request), so the formatter adopted next keeps every
function within 40 lines; three complexity baseline entries are gone. Behaviour unchanged. - Internal: every refusal that quotes a Zod error's issues (
path: message) is built by one helper,src/issues.ts
(seven inline copies in six files before). Messages unchanged. - Development only: Prettier formats the code (
npm run format;npm run lintchecks it, in CI too), printWidth 120 —
the style the code already had; the code reformatted once, in three pull requests that each fit one review
(scripts/,src/,test/);.git-blame-ignore-revsnames all three. - Runtime dependencies:
@modelcontextprotocol/sdk1.32.1 (was 1.32.0) andzod4.6.7 (was 4.6.5). - Development only:
@babel/coreand@babel/eslint-parser8.0.7,jscpd5.4.1,prettier3.9.10 (the code's
formatting is unchanged), andactions/setup-node7.1.0 in the workflows.
2.5.0 — image families, URL inputs, account, training, video, commercial tools
One release for steps 1–5 of docs/DESIGN-ideogram-v2.md: the paragraphs below state each step's counts as they stood after that step; the release ends at 38 tools (tools/list 131 396 bytes, budget 136 KB), support curated 80 · raw 47 · planned 0 · unsupported 73, 178 tests. Nothing a 2.0.0 client used is removed (the video input of ideogram_generation existed only inside this series).
Step 1 of docs/DESIGN-ideogram-v2.md: every image family of Ideogram's documentation index has its own tool.
Step 2: the account — usage and spend, invoices, API keys — and the webhook signature helper.
Step 3: custom-model training — datasets, their assets, the training of an Ideogram 4.0 or 3.0 model, the models.
Step 4: video — from a prompt, from an image, from references.
Step 5: the commercial tools and the v1 utilities, each one operation; nothing of the specification is planned any more.
Added
- A file field (image, mask, reference and style images, fonts) takes a public
https://URL beside a local path, on
every tool (the family tools,ideogram_quote,ideogram_api): fetched one after another under the field's own
limit and the room the request cap leaves, as a type the field takes, from a public host only — no IP literal, no
local name, no private / loopback / link-local / carrier / site-local / mapped / translated address after
resolution, no redirect followed, never an empty body — inside the call's budget, over a connection pinned to the
very address the lookup answered (a DNS rebinding reaches nothing), local files read before any fetch, every
failure named with the URL and its cause; the threat model is section 9b of the design note. inline_images: trueon every tool that saves images (the family tools,ideogram_generation,ideogram_api):
each saved image under 3.75 MB (the model APIs behind the clients take 5 MiB of base64 per image), up to 10 MB per
result, is also returned as MCP image content with its media type normalized, for a client without access to this
machine's files; the others are named with their size and stay by path. Off by default.ideogram_usage: the organization's billed API usage as Ideogram reports it (GET /v2/account/usage: dense time
buckets of line items — product, endpoint, cost, billed units, the redacted API key, the source), the last 7 days
by day (24 hours by hour) unless told otherwise, a range over the API's span (92 days at1d, 168 hours at1h)
refused before asking, summed per product, unit and currency as decimal strings (never floats; an amount the sums
cannot read is counted and listed, not summed — an unreadable quantity leaves the cost in), the buckets written
exactly as received to a JSON file in the output directory (named by the whole query, owner-readable: the line items
carry emails and key prefixes) and printed too when under 64 KB: the shape ai-cost reads; the end of the range is
always sent, so the span the tool checked is the span the API sees; every account answer — the invoices, the API
keys, one off its specification (reported as such, never as a success) — is written to an owner-readable file the
same way and shown inline under 64 KB.ideogram_invoicesand
ideogram_api_keyslist what their endpoints return as received. All three need an organization-admin key: a 404
is said as that, with Ideogram's words. Reads only, never billed.verifyWebhook(dist/webhooks.js): the canonical message Ideogram signs
(request_id\nuser_id\ntimestamp\nsha256_hex(body)) checked against the JWKS ofGET /v1/.well-known/jwks.json
with Ed25519, the header's key first, a rotated key still accepted, the signature as base64, base64url or hex; a
replay refused by the timestamp (300 s by default,Infinityturns it off; the signature alone never expires); a
missing header is null, never a throw. A helper for the receiver: this server receives no webhook itself (a stdio process has no public
URL); whererequest_id,user_idandtimestamptravel is not stated by the specification.ideogram_operationsand the API reference list an operation's path and query parameters.ideogram_datasets(the list, searched by name, or one dataset with its files, captions and the models trained
from it),ideogram_dataset_upload(images, .txt caption sidecars and .zip archives — local paths or public https
URLs — into an existing dataset or a new one created from a name; what was accepted and what failed, with the
reason),ideogram_train(an Ideogram 4.0 or 3.0 model from a dataset of 15 to 100 images; any hyperparameter
routes to the advanced operation, each checked by its own schema; a field the model's training does not take is
refused by name) andideogram_models(the list by scope and status, or one model with its training runs): the
v1 endpoints the index keeps for training, curated.train_dataset_modelis served raw. Support: curated 57 · raw
6 · planned 64; tools/list: 23 tools.- A dataset created by
ideogram_dataset_uploadis named first in every answer that follows (an upload error, a
200 off the schema, a file that cannot be written), so a retry uploads into it instead of making another; a create
answered off its schema says a dataset MAY exist. A URL's last segment is percent-decoded before it names the part
(red%20door.pngpairs with its caption asred door.png). every call with a side effect (create a dataset, upload assets, start a
training) has one typed outcome (sendOperation): refused, not sent, unknown, or the answer — a refused or never-sent
upload says "Nothing was uploaded", an unknown one "the assets MAY be in the dataset", an unknown train "a training
MAY have started and will be billed"; the client's status-0 errors carrydetails.sent(false when the budget ended
or the connection was refused before anything left this machine). The raw answer's file is best-effort: a directory
that cannot be written shows the answer inline and never turns a success into an error (rawWriteFailurescounted);
a LISTING whose answer is neither written nor small enough to print is an error (the data is the listing). The same
typed decision reaches every surface that sends a request with a side effect: a family tool orideogram_apiwhose
non-GET request fails after it left (a 5xx, a cut connection) says the operation MAY have run and be billed. A
decoded URL segment is cut at its last slash or backslash before it names a part (..%2F..%2Fx.pngisx.png). ideogram_video_text,ideogram_video_image,ideogram_video_reference: the three video families on the family
mechanism (Seedance 2.5 by default, Seedance 2, Kling 3 Standard, MiniMax H3; each model its own fields), async
only — a finished video is saved as mp4 (under its own 512 MiB download cap) with its measured length, resolution
and size, never returned inline (the video tools do not advertiseinline_images; asked anyway, the answer says it
has no effect); a long job comes back as a generation_id forideogram_generation. Audio references
(reference_audios) are sent as MP3 or WAV — the one field-type rule, anchored to the specification; a start and
an end image as parts or assets. The video families show their own sentences for the fields they share with the
image families (a resolution is a tier and goes with the aspect ratio; references are not an edit's) through one
lookup,fieldTextsFor. Rules anchored to the specification's prose: a reference-to-video model (MiniMax included)
needs a reference image or video;reference_images/reference_image_asset_identifiers,image/
image_asset_identifierandend_image/end_image_asset_identifierare alternatives; MiniMax takes at most 12
references across images, videos and audio.durationsays what an omitted value costs: Kling and MiniMax make
5 s, Seedance picks the length and is QUOTED at its maximum (measured 2026-10-11: seedance-2-5 omitted and 30 s
6.93 USD, 4 s 0.924, 10 s 2.31 at 720p; the dry run does not pricegenerate_audio;docs/PRICES-2026-10-11.md
quotes every video model at its minimum duration — Kling 3 s 0.252, Seedance 2 4 s 0.64). A saved image's line names
its Ideogram asset id, for the*_asset_identifierinputs of a later call. The spec-only video edit is served raw
behindallow_undocumented. Support: curated 68 · raw 7 · planned 52; tools/list: 26 tools.ideogram_generationlost itsvideoinput: the poll gaps within a 50 s wait (2, 3, 4.5, 6.75, 10.1, 15.2 s)
never reach either cap, so the knob had no effect; one cap for images and video. An empty result list says "no
image or video". A download the budget cuts names the re-collect path for videos as for images (one download loop).- A download is checked against the kind its result promised: a video link that answers an image (a placeholder behind
a 200) is "Not saved: Expected a video", never counted as a saved video; an image link that answers a video likewise.
A saved image'sAsset:line is the whole identifier ({"asset_type": "RESPONSE", "asset_id": …}), ready for an
*_asset_identifierinput.ideogram_quote(andideogram_apiwithdry_run) says when the dry run leaves
generate_audioout of the price (measured: the same quote with and without it; Ideogram: about double). A video
entry that carries adurationbut noobject_typeis a video (the specification leaves the type optional). get_asset_reference_usage(spec-only) is served byideogram_apibehindallow_undocumented: a spec-only
operation of a curated family is raw, as the undocumented reframe models are. Support: curated 47 · raw 5 ·
planned 75. tools/list: 19 tools.- Twelve single-operation tools (
src/tools/commercial.ts) on the family mechanism — aFamilyToolSpecmay name one
operation: nomodelinput, the operation's own fields, the same checks (schema, the shared rules, `pr...
2.0.0 — the Ideogram platform on the v2 API
The server moves to Ideogram's v2 API and is built from its OpenAPI specification: every image model the platform
sells through an API key is reachable, priced before a call, and collected by id when it runs long
(docs/DESIGN-ideogram-v2.md, docs/adr/0001-openapi-snapshot-as-the-source-of-truth.md).
Added
spec/openapi.json: the specification snapshot of 2026-10-07 (200 operations), classified by one rule into
documented 66 / spec_only 20 / v1_only 41 / legacy 32 / internal 9 / bearer_only 32; the types and zod schemas
generated from it (src/generated/,@hey-api/openapi-ts), regenerated and compared in CI.modelon every family tool, with the exact fields of each model (one variant per model in the advertised schema).
ideogram_generatereaches 18 models: auto, Ideogram 4.5, 4.0 (+ custom model, transparent), 3.0 (+ character,
custom model, transparent), 2a, 2.0, GPT Image 2 / 2.5 Flare / 2.5 Sunburst, Nano Banana 2 / Pro, P-Image, Z-Image.
Remix adds Ideogram 4.0 and auto; reframe Nano Banana 2; replace background GPT Image 2; upscale the Topaz models,
Nano Banana Pro and auto; describe Ideogram 4.0 (a structured JSON prompt). A field the chosen model does not take is
refused, naming the models that take it.ideogram_inpaint(the v2 name ofideogram_edit, which stays as its alias through 2.x).ideogram_quote: the price of exactly the call a tool would make, from the API's owndry_run(USD and credits as
decimal strings, exact or an estimate with its upper bound; nothing generated or billed). Operations without
dry_run(describe) are refused locally — a quote request could run them.ideogram_generation: collect a generation by id. Every generation is sent asynchronously and returned at
acceptance; the tool waits up towait_s(default 45, at most 50 — the MCP client times out at 60) and otherwise
answers with the id.ideogram_operations(discovery) andideogram_api: any operation this release serves, by id, with its path,
query, header and body parameters checked against the operation's generated schemas and the reviewed constraints;
undocumented operations needallow_undocumented: true; legacy, internal and Bearer-only operations are refused.- Typed 402 / 429 errors (
reject_reasonand what to do about it,Retry-After, the in-flight limit). - A weekly drift check of the live specification (
scripts/spec-pull.mjs,.github/workflows/spec-drift.yml): one
issue, updated, never a failed build. docs/API-REFERENCE.mdgenerated from the snapshot (family × model × fields).
Changed
- Every call goes to
/v2/…; images are downloaded by streaming to disk with a byte counter (50 MB cap) instead of
buffering whole bodies. - Upload limits are each operation's own, as the specification states them (describe 10 MB, remix 50 MB, 4.5's
images 25 MB × 5, …), checked before any file is read ("MB" read as decimal megabytes); JPEG, PNG and WebP, the types the specification names. - A request is sent again only when it never left the machine or the API rejected it with 429; a 5xx, a timeout or a
reset after sending is reported, never repeated (one tool call never creates two billed jobs). Polls and downloads
(GET) keep retrying network failures, 429 and 5xx. - One budget per tool call: 55 s (the MCP client's 60 s minus a margin), ended earlier by the caller's cancellation;
every attempt, retry sleep (Retry-After included), poll and download of the call is judged against what remains, so
nothing of a call outlives its caller — a POST is resent only with time for an attempt as long as the rejected one,
a cut download is listed as not saved beside the generation id (typedCALL_TIMEOUT/CANCELLED), and
ideogram_apirefusesasync: false(a synchronous result this server could not wait for). - A cut attempt is typed by construction (review rounds 3–6 of #36, two recorded step-backs): one timer per bound — the
budget's own signal alone when it is the shorter bound, so no second timer races the same instant — and a poll the
wait cuts in flight is a clean Pending (collect it withideogram_generation), never "the last poll failed". - A 2xx body the specification's schema rejects is reported as a contract mismatch with the generation id, never as a
success.
Migration from 1.x
The 1.x tool names keep working; a 1.x call is mapped and the result says what was mapped.
| 1.x | 2.0.0 |
|---|---|
ideogram_generate default (Ideogram 3.0) |
the same model, ideogram-3 (cost-neutral; ideogram-4-5 is the newest) |
model: "3.0" / "4.0" |
ideogram-3 / ideogram-4 (aliases through 2.x) |
rendering_speed: "TURBO" | "DEFAULT" | "QUALITY" |
lowercase on the models that take rendering_speed (Ideogram 2.x, 3.x, 4.0); quality: "low" | "medium" | "high" on those that take quality (Ideogram 4.5, GPT Image 2.5, P-Image) |
rendering_speed: "FLASH" |
refused where rendering_speed is taken (v2 has no FLASH); quality: "very_low" where quality is (GPT Image 2.5 has no very_low: refused there by its schema) |
magic_prompt, style_type uppercase; color_palette.name: "EMBER" |
lowercase (auto, realistic, ember) |
ideogram_edit |
ideogram_inpaint (the alias stays through 2.x) |
ideogram_describe describe_model_version: "V_3" |
model: "ideogram-3"; "V_2" is refused (no v2 model) |
character_reference_image (one file) on generate / remix / edit |
model: "ideogram-3-character" with character_reference_images |
custom_model_uri on generate |
model: "ideogram-3-custom-model" or "ideogram-4-custom-model" |
ideogram_upscale (default model auto) prompt / detail |
taken by other upscale models (prompt: nano-banana-pro, topaz-bloom-2, topaz-redefine; detail: topaz-redefine) — refused on auto, naming them |
ideogram_upscale resemblance / magic_prompt |
no v2 upscale model takes them: refused; upscale_factor sets the size |
ideogram_replace_background magic_prompt / seed |
not taken by v2's replace-background models: refused by name |
ideogram_generate with model 4.0 refusing 3.0 fields |
each model's own fields, refused naming the models that take them |
webhook_url, private and target_collection_id are reachable through ideogram_api; the curated tools poll and
save locally instead.
Removed
- The v1 client, the hand-written schemas of the 1.x tools and their endpoint tests.
v1.2.1
Added
.github/workflows/mcp-registry.ymlkeeps the official MCP Registry in step with the releases: every run (at a
release, at the end of the Release workflow, daily, by hand) publishes each of the last 10 releases that npm
serves and the registry lacks, oldest first — theserver.jsonof its tag with the default branch's description
(GitHub Actions OIDC, no secret;mcp-publisherpinned by version and sha256). The registry listed an old version
of this server; the next run lists the missing ones.
Fixed
server.jsondescriptionwithin the registry's 100-character limit (the registry refused the longer one).
Security
@modelcontextprotocol/sdk1.32.0 (was 1.30.1), past GHSA-6qxp-vccf-f47h (fixed in 1.31.0: the SDK's OAuth client
could send credentials to an authorization server the MCP server chose). This server runs over stdio and imports only
the SDK's server side, so it never used that client; the bump clearsnpm audit.
v1.2.0
A comparison with Ideogram's own MCP server and the current API (docs/COMPETITION-2026-10-01.md).
Added
- Style controls on
ideogram_generate(3.0),ideogram_editandideogram_remix:style_reference_images(1-3 files),
character_reference_imagewith an optionalcharacter_reference_mask,style_codes(1-8),style_presetand
color_palette(a preset name or 1-10 explicit colours with weights, never both) on all three; an exactresolution
(one of the 69 sizes, not withaspect_ratio) on generate and remix;custom_model_uriand
enable_copyright_detectionon generate — each endpoint gets exactly the parameters its documentation lists, sent
under the API's field names (a list as repeated fields, the palette as oneapplication/jsonpart as the OpenAPI
spec declares — a live probe on 2026-10-01 accepted it; the tests stubfetch). The three tools' input schemas
are strict: a parameter an endpoint does not take (an exactresolutionon edit) is refused by name, never
silently dropped. Every file
counts against the 50 MB request limit, checked withstatbefore any file is read. Tests pin the encoding, the
per-endpoint fields and the refusals on all three endpoints. ideogram_generatewithmodel: "4.0"posts to/v1/ideogram-v4/generate(text_prompt,resolution,
rendering_speedexceptFLASH,enable_copyright_detection); a 3.0-only parameter given with 4.0 is refused by
name, never dropped. A live probe on 2026-10-01 returned the documented shape (response_type,created,
data[].url/prompt/resolution/is_image_safe/seed), which the response schema accepts.
Changed
- README and
docs/API-REFERENCE.mdsaid "Ideogram 4.x uses the v2 API": the v1 API has theideogram-v4endpoints too.
The v2 API (Ideogram 4.5, Precise Edit, async jobs, usage) is still not called.
v1.1.0
An audit of the npm page (2026-10-01, every claim checked against the code and Ideogram's documentation).
Changed
ideogram_editposts to/v1/ideogram-v3/inpaint: Ideogram marks/v1/ideogram-v3/edit"Legacy: use inpaint instead. This endpoint will be removed in a future release" — same image, mask and prompt fields, same response. A test with a fakefetchpins the endpoint.- The per-image cap is 25 MB, Ideogram's documented maximum (it was 10 MB, called an API limit); the error names the real limit. A test pins both sides of the cap.
ideogram_editrefuses image + mask at or over Ideogram's 50 MB request limit before uploading anything. A timed-out POST is no longer retried: the server may have accepted and billed it, and the multipart body (now up to 50 MB) would be sent again; connection failures and 429/5xx are retried as before, and a timed-out image download (an idempotent GET) still is. The pair is measured withstatbefore either file is read, with 64 KB reserved for the multipart overhead. TheNETWORK_ERRORmessage names the failure (TimeoutError,TypeError). - Every public claim says what the code does: Ideogram 3.0 through the v1 API (Ideogram 4.x and the v2 API are not used; the v1 API keeps working with no announced sunset); the describe and upscale endpoints named as such;
style_typedefaults to GENERAL on the API side (the descriptions said AUTO);num_imagesandseedin every tool exceptdescribe;describe_model_version,negative_prompt, upscale'spromptandimage_weightdocumented; API keys at ideogram.ai/platform (the old manage-api link answers 403); Node.js ≥ 22 stated; theideogram-mcpcommand and aclaude mcp addone-liner;npm testandnpm run typecheckin Development; the mask has no "10 % black" rule; SECURITY.md labels the review history (1.0.0, April 2026) and says a symlinked parent directory is followed and that any readable image path the model names may be uploaded; "zero-trust" and "flagship" dropped; the npm and MCP registry descriptions agree. package.jsonnames the repository asgit+https://…— the form npm publishes, so a publish prints no auto-correction.
v1.0.4
Changed
- Runtime dependencies:
@modelcontextprotocol/sdk1.30.1 (was 1.30.0) andzod4.6.5 (was 4.3.6). - Releases are staged on npm instead of published directly: the release workflow runs
npm stage publish
(Trusted Publishing, provenance), and a version becomes public only when a maintainer approves it on
npmjs.com with two-factor authentication; the workflow pins npm 11.20.0 for it (staged publishing needs 11.15.0+). - The README opens with the Quantum Media Technologies wordmark and closes with a "Made by" line, both linking to www.qmediat.io/open-source;
package.jsonhomepagepoints there (authoralready carried the company line). - Development: the project builds with TypeScript 7 (the native compiler). The published JavaScript is byte-identical
to 1.0.3; the type declarations describe the same types (only the order of union members and properties and the
quote style differ) and the source maps are regenerated. The TypeScript 7 npm package ships onlytsc— no
tsserverand no JavaScript API — so an editor set to use the workspace TypeScript version cannot load it from
node_modules; use the editor's bundled one. - The release workflow checks that
CHANGELOG.mdhas a section for the version, asks npm once whether the version
is already public and stops on any answer other than yes or no (a registry error used to read as "not public"),
runs an npm pinned by version and by the sha512 of its tarball, and names a failed stage in the run summary.
v1.0.3
Security
- Lockfile refresh closes all 45 open Dependabot advisories (9 high; all transitive dependencies of the MCP SDK):
hono4.13.8,fast-uri3.1.8,qs6.16.0,body-parser2.3.0,ip-address10.7.2.npm auditreports
0 vulnerabilities.@modelcontextprotocol/sdkis pinned to 1.30.0, the version this release was tested with;
zodstays at 4.3.6. The advisories sit in transitive dependencies under the SDK's caret ranges, so a project
that already has this package in its lockfile keeps its old tree until it runsnpm update(ornpm audit fix);
a fresh install gets the refreshed tree. - CI now typechecks, runs the smoke test and refuses a build with a known high-severity advisory in the shipped
dependency tree (npm audit --omit=dev --audit-level=high); the CI token is read-only. - The publish workflow verifies that the release tag equals the
package.jsonversion before publishing and runs
the same audit gate. - Dependabot configuration (
.github/dependabot.yml): weekly grouped npm updates and GitHub Actions updates. - Dependabot security updates and GitHub private vulnerability reporting are enabled on the repository;
SECURITY.mdpoints to the private reporting form first.
Fixed
- The server advertised
version: 1.0.2inserverInforegardless of the package version; it now reads the
version frompackage.json, and the smoke test asserts the two agree. - The publish workflow accepted a release tag without the
vprefix; the contract isv<version>. - Starting without
IDEOGRAM_API_KEYprinted a serialized Zod issue list namingapiKey; it now prints
Configuration error: IDEOGRAM_API_KEY is required(found by the new smoke test).
Added
npm test(builds first): a stdio smoke test on Node's built-in runner — the built server completes the MCP
handshake, advertises the package version and lists its seven tools; it refuses to start without
IDEOGRAM_API_KEY. Every request has a 10 s deadline and is rejected if the server exits. No new dependency.
v1.0.0 — Initial Release
ideogram-mcp v1.0.0
MCP server for the Ideogram V3 API — 7 tools, security-first, 2 runtime dependencies.
Tools
ideogram_generate— text-to-image generationideogram_describe— image-to-text descriptionideogram_edit— mask-based inpaintingideogram_remix— image transformation with style controlideogram_reframe— outpainting to new resolutions (69 valid sizes)ideogram_replace_background— foreground-preserving background replacementideogram_upscale— guided image upscaling
Install
npm install -g @qmediat.io/ideogram-mcpSecurity
Zero axios, SSRF protection, symlink rejection, Zod response validation, Content-Type enforcement, retry with backoff. Full details in SECURITY.md.
Quality
6 rounds of code review by GPT-5.3-Codex, Gemini 2.5 Pro, Grok-4, and GitHub Copilot. 49 issues found and fixed. All 7 tools live-tested with real Ideogram API.
Built by Quantum Media Technologies