Ship compliance features. We maintain the framework knowledge.
The compliance router: one endpoint, 150+ framework modules, routed per request. Curated compliance modules injected at inference, disclosed per response.
Use the ISMS Copilot API when your own code needs compliance answers: ISMS Copilot maintains curated modules for 150+ frameworks, and the response names the modules that ran.
https://api.ismscopilot.com/v1
curl https://api.ismscopilot.com/v1/chat/completions \
-H "Authorization: Bearer sk-isms-..." \
-H "Content-Type: application/json" \
-d '{
"model": "isms-fast",
"messages": [
{"role": "user", "content": "List 3 ISO 27001 access control requirements."}
]
}'TL;DR
ISMS Copilot APIOpenAI-compatible text-only /v1/chat/completions
Any foundation-model APITokens from a general-purpose model; wire varies by provider
ISMS Copilot APICurated modules, detected or pinned, injected at inference
Any foundation-model APITraining recall plus your system prompt
ISMS Copilot APIISMS Copilot. Public GET /v1/frameworks catalog
Any foundation-model APIYou own the corpus forever
ISMS Copilot APIx-isms-frameworks header names the modules that ran
Any foundation-model APINo built-in module list on the response
ISMS Copilot APIauto, none, or pin up to 8 catalog ids
Any foundation-model APIRoll your own retrieval
ISMS Copilot APIDocumented Zero Data Retention for prompts and outputs
Any foundation-model APIOften retained for a limited abuse window
The API in 27 seconds
API or MCP server: pick by who makes the call
APIYour own code: your product, CI, scheduled jobs, pipelines
MCP serverYou, inside your AI tool: Claude, ChatGPT, Cursor, Codex or any MCP client
APIAn sk-isms key and a new base URL in your OpenAI SDK
MCP serverThe server URL, then sign in (Claude, ChatGPT) or a personal access token (Cursor, Codex)
APIPrepaid API credits, billed per token
MCP serverYour ISMS Copilot chat plan, including Free
What comes back
Trimmed. The x-isms-frameworks header and the ismscopilot object tell you which curated modules were injected and how much knowledge went in.
HTTP/1.1 200 OK
x-isms-frameworks: ISO_27001
{
"model": "isms-fast",
"choices": [
{ "message": { "role": "assistant", "content": "ISO/IEC 27001:2022 Annex A, control A.5.15 ..." } }
],
"usage": { "prompt_tokens": 1842, "completion_tokens": 312 },
"ismscopilot": {
"frameworks": ["ISO_27001"],
"knowledge_characters": 4926,
"knowledge_tokens_method": "chars_div_3_v1"
}
}Same model, with and without the knowledge
On this frozen 20-task compliance set, the same base model scored differently depending on the scaffold around it; the largest increase came when curated framework knowledge was injected at inference time, especially on less-known frameworks.
- Naked
- 68.0% overall
- Best DIY prompt
- 76.0% overall
- DIY + knowledge module
- 96.9% overall
- Full ISMS Copilot
- 92.1% overall
The module is the same curated reference ISMS Copilot injects when your question names a framework, and the same one the API exposes. Outside the trap questions, the naked model fabricated 8 control and article identifiers, including an entire invented ISO 42001 Annex A structure. The product fabricated zero. The primary endpoint was frozen in advance: full product versus the better of the two DIY prompts. The pre-registered verdict is a tie, and that is the verdict we report.
Try the catalog without a key
No key needed. Returns count and registry_hash. 150+ modules; the live catalog is authoritative.
curl https://api.ismscopilot.com/v1/frameworksModels
The four standard aliases cost $2.80 per million input tokens and $8.80 per million output tokens, on either path. isms-mini costs $0.50 in and $2.00 out. The console stays the billing source of truth.
- isms-mini
- GLM-5.3-Flash, global, bulk lane, no thinking mode
- isms-fast
- GLM 5.3, global
- isms-thinking
- GLM 5.3, global, reasoning
- isms-fast-eu
- GLM 5.3 hosted by Mistral, EU processing path
- isms-thinking-eu
- GLM 5.3 hosted by Mistral, EU processing path, reasoning
Framework control
auto detects named frameworks in the conversation, none injects nothing, or pin up to 8 catalog ids. ismscopilot.system: "none" opts out of the persona.
{
"model": "isms-thinking",
"messages": [{ "role": "user", "content": "Map A.5.23 to SOC 2." }],
"ismscopilot": {
"frameworks": ["ISO_27001", "SOC_2"]
}
}One product, customers in many countries
Building a compliance product with its own assistant? Pin each customer's national modules on their requests: NIS_2_NL for a Dutch customer, NIS_2_IT for an Italian one. Every EU member state has its own NIS 2 module (tracking the draft where the national law is still pending), and each request carries only its own, so adding a country never changes answers for the customers you already have.
The response names the modules that ran. Changes to any module are listed at GET /v1/frameworks/changelog.
curl https://api.ismscopilot.com/v1/chat/completions \
-H "Authorization: Bearer sk-isms-..." \
-H "Content-Type: application/json" \
-d '{
"model": "isms-fast",
"messages": [
{"role": "user", "content": "Which NIS 2 incident reports do we owe, and to whom?"}
],
"ismscopilot": { "frameworks": ["NIS_2_NL"] }
}'Auth and limits
- Key
- sk-isms-... shown once, created in the console
- Header
- Authorization: Bearer sk-isms-...
- Keys per user
- Up to 20 active, optional spend caps per key by hour, day, week, month
- Streaming
- Supported on both processing paths
- Not supported
- Tool calling, JSON mode, n > 1, logprobs, images. Text-only subset.
- Thinking floor
- max_tokens raised to 1024 on thinking aliases; effective value returned in x-isms-max-tokens-effective
Credits
- Billing
- Prepaid credits, no subscription
- First-key credit
- USD 2 once, which a new API account can get when it creates its first key in the console (one per person; accounts that sign in only with Microsoft do not qualify; while the offer runs)
- Minimum top-up
- USD 20
- Welcome credit
- USD 10 once, for paying chat subscribers
- Unit rates
- platform.ismscopilot.com/pricing
Zero Data Retention
Prompts and outputs are not stored in the API data layer. Console usage history is metadata only. The global path runs on zero-retention upstream endpoints; the EU path runs on Mistral under a commercial ZDR agreement. Request content is never used for training. Usage and billing metadata are retained.
Works with
- Any OpenAI SDK
Swap base_url and the key. Python, Node, Go, Ruby, the rest.
- opencode, Continue, Aider, Cline
Custom OpenAI-compatible (text-only) provider pointed at the base URL.
- CursorChat only
Plain chat only, through the Override OpenAI Base URL setting (caveated, not a verified integration). Agent mode sends tool calls this API rejects: use the sub-agent step.
- LangChain, Vercel AI SDK, n8n, LiteLLM, Cloudflare AI GatewayGuides
Setup guides for each, with framework pinning and the text-only limits stated.
- Claude CodeSub-agent
Cannot use this endpoint as its model backend (Anthropic API shape). Run it as a compliance sub-agent step, or use Account MCP.
Related reading
Status: live
Create an sk-isms key at platform.ismscopilot.com/keys and send the first request. While the offer runs, a new API account can get a one-time USD 2 credit when it creates its first key in the console. It is one per person, and accounts that sign in only with Microsoft do not qualify. Machine summary: GET /api/public/api/v1.
Frequently Asked Questions
Is it fully OpenAI-compatible?
It is a documented text-only subset of the chat completions API: no tool calling, no JSON mode, n=1, no logprobs, no images. Streaming works. Existing OpenAI SDKs work once you swap the base URL and the key.
Does it search the standard at request time?
No. Detection is name-level: a named framework in the conversation selects a curated module that is injected before generation. A bare clause number injects nothing. No verbatim ISO or AICPA text is returned.
Which frameworks?
150+ curated modules. GET /v1/frameworks returns the live list with a registry hash and needs no key.
What does it cost?
While the offer runs, a new API account can get a one-time USD 2 credit when it creates its first key in the console. It is one per person, and accounts that sign in only with Microsoft do not qualify. After that, prepaid credits with a USD 20 minimum top-up. Paying chat subscribers get a one-time USD 10 welcome credit instead. Unit rates: $2.80 in and $8.80 out per million tokens for isms-fast, isms-thinking and their -eu twins; $0.50 in and $2.00 out for isms-mini. The console at platform.ismscopilot.com/pricing is the billing source of truth.
Where does it run?
Global aliases run through zero-retention upstream endpoints; the -eu aliases run on Mistral in the EU. The processing region comes back in a response header.
Can I use it from Claude Code?
Not as the model backend; Claude Code speaks the Anthropic API. Use it as a sub-agent step from a script, or connect Account MCP for account-level access.
Send the first request tonight
Create a key and paste the curl. New to the API? While the offer runs, a new API account can get a one-time USD 2 credit when it creates its first key in the console. It is one per person, and accounts that sign in only with Microsoft do not qualify. The response tells you which modules ran.
ISMS Copilot