API reference
Every endpoint of the ProvenDecisions API (version 1) and every tool of the MCP server, with an example call and a real answer from our own copy of the site, shortened. Start with the Developers page for keys and setup. Machine-readable: OpenAPI document and endpoint list.
Basics: address, key, versions · Plans, limits and paging · Errors · Solving · Content · MCP tools · OpenAPI document
Basics
- Address:
https://proven-decisions.ai/api/v1/...; answers are JSON (UTF-8). - Key: send
Authorization: Bearer pd_live_...with every request. Make keys in the app under Settings, API keys (shown once; we keep only a scrambled copy). - Or an OAuth sign-in (for claude.ai custom connectors and other MCP clients that sign in): discovery at
https://proven-decisions.ai/.well-known/oauth-protected-resourceand/.well-known/oauth-authorization-server, dynamic client registration at/oauth/register(public clients), the authorization code grant with PKCE S256 at/oauth/authorizeand/oauth/token, refresh tokens (replaced at each use) and/oauth/revoke. Access tokens last 1 hour, refresh tokens 30 days. The access token goes in the sameAuthorization: Bearerheader and works for/mcpand/api/v1/*with the same plans and limits as a key. A 401 from/mcpnames the discovery document inWWW-Authenticate. - Plans: Standard and Pro. Free and Student plans stay web-only (403).
- Solving takes typed math or LaTeX only (the same as the Math box in the app); word, photo and voice problems are solved in the web app. No language model runs behind the API: every number comes from a solver and is checked.
- Versions: this is v1. Within v1 we only add things (new endpoints, new fields in answers); programs should ignore fields they do not know. A change that would break a program gets a new version (
/api/v2), and v1 keeps working for a while after it, with notice on this page.
Plans, limits and paging
Requests from the API, the MCP server and the website count together for your account:
- Solving (solve, dual, kkt, verify and the MCP tools solve, dual, kkt, explain): at most 5 requests a minute, 100 a day and 1,000 a month; one problem at a time; 10 minutes of solving a day; up to 3,000 characters a problem; the monthly stop at 85% of your plan's price.
- Content (search, tutorials, examples, Daily Update, posts, Help, reports and the MCP content tools): the lighter limit the website uses for searches, 30 requests a minute and 1,000 a day. They do not use up your solves.
The website's plan rules apply. Tutorials above your plan are listed with "locked": true and are not served (403, code plan_locked): a Standard key reads Free, Student and Standard tutorials, a Pro key all of them. The calculus Refreshers are free. The Daily Update covers the last 365 days, as for a signed-in user on the website; earlier periods are reviewed with Update On Demand in the app. Drafts are never served.
Paging: lists answer {kind, total, limit, offset, next_offset, items}. Ask for the next page with offset=next_offset until next_offset is null.
Errors
Every refusal is JSON {"detail": "...", "code": "..."}; detail says what to do in plain words.
| HTTP | code | When |
|---|---|---|
| 400 | empty, words_not_supported, typed_math_only, bad_kind, bad_paging, bad_date, bad_section, bad_format | A parameter is missing or not valid; a word problem sent to solve. |
| 401 | unauthorized | No key, or an unknown or revoked key. |
| 403 | plan | Your plan has no API access, or vector syntax on a plan below Pro. |
| 403 | plan_locked | A tutorial above your plan. |
| 404 | not_found, out_of_window | No such item, or a Daily Update older than your window. |
| 413 | A model longer than the limit. | |
| 422 | parse_error, no_dual, no_kkt | The model could not be read, or it has no dual or KKT certificate. |
| 429 | limit, monthly_stop | A limit is reached; the message says which and when it resets. |
Solving
POST /api/v1/solve
Solve a model: the solver's status, the objective, the solution, the check of every constraint, the dual or KKT certificate, the proof label and the walkthrough steps. The product mix example: make products A and B to earn the most profit ($) within 100 labour hours and 80 units of material.
| Parameter | Where | What |
|---|---|---|
model | body | Typed math or LaTeX, up to 3,000 characters. |
tier | body | Low, Medium (default) or High. |
depth | body | Walkthrough depth 1 to 10 (default 3). |
test_dual | body | Add the dual or KKT certificate (default true). |
Example:
curl -s "https://proven-decisions.ai/api/v1/solve" \
-H "Authorization: Bearer $PD_KEY" \
-H "Content-Type: application/json" \
-d '{"model": "maximize 3*a + 2*b\nsubject to\n2*a + 1*b <= 100\n1*a + 2*b <= 80\na, b >= 0"}'Answer (shortened):
{
"status": "optimal",
"objective": 160.0,
"solution": {
"a": 40.0,
"b": 20.0
},
"proof_label": "Exactly proven optimal",
"verification": {
"feasible": true,
"max_violation": 0.0,
"verified": true
}
}The full answer also has constraints (slacks), dual, proof, model_text, latex, steps and wall_ms.
POST /api/v1/dual
The dual of a linear program or convex quadratic program, solved, with the strong or weak duality report.
| Parameter | Where | What |
|---|---|---|
model | body | Typed math or LaTeX. |
tier | body | As above. |
Example:
curl -s "https://proven-decisions.ai/api/v1/dual" \
-H "Authorization: Bearer $PD_KEY" \
-H "Content-Type: application/json" \
-d '{"model": "maximize 3*a + 2*b\nsubject to\n2*a + 1*b <= 100\n1*a + 2*b <= 80\na, b >= 0"}'Answer (shortened):
{
"model_text": "minimize 100*y1 + 80*y2\nsubject to\n y1 >= 0\n y2 >= 0\n 2*y1 + y2 >= 3\n y1 + 2*y2 >= 2",
"status": "optimal",
"objective": 160.0,
"solution": {
"y1": 1.3333333333333333,
"y2": 0.3333333333333335
},
"duality": {
"kind": "strong",
"primal_status": "optimal",
"dual_status": "optimal",
"gap": 0.0,
"holds": true,
"explanation": "Strong duality: primal and dual optima are equal (160)."
}
}POST /api/v1/kkt
The KKT (Lagrange multiplier) certificate at the solution of a smooth model: the multipliers, the residual and whether the conditions hold. POST /api/v1/verify is the same.
| Parameter | Where | What |
|---|---|---|
model | body | Typed math or LaTeX. |
tier | body | As above. |
Example:
curl -s "https://proven-decisions.ai/api/v1/kkt" \
-H "Authorization: Bearer $PD_KEY" \
-H "Content-Type: application/json" \
-d '{"model": "minimize (x - 1)^2 + (y - 2)^2\nsubject to\nx + y <= 1"}'Answer (shortened):
{
"applies": true,
"holds": true,
"residual": 4.440892098500626e-16,
"multipliers": {
"constraint 1": 1.999999999999996
},
"global": true,
"note": "KKT conditions hold and the problem is convex, so the Lagrangian dual has no gap: the answer is globally optimal."
}Content
The site's own content, read-only: tutorials and calculus refreshers, solved examples, the Daily Update (optimization news), New in AI (a monthly post), Help articles and Update On Demand reports. Tutorials come as Markdown with the math kept as LaTeX ($...$ inline, $$...$$ on its own line). Each item has url (in the API) and page_url (on the website).
GET /api/v1/content/search
Search everything at once, with the website's own search for each kind; best first within each kind. To keep your search words out of web server logs, send the same fields as JSON with POST /api/v1/content/search.
| Parameter | Where | What |
|---|---|---|
q | query | Search words. |
kinds | query | Comma list, default all: tutorials, refreshers, examples, daily, posts, help, reports, glossary. |
limit | query | Hits per kind, 1 to 20 (default 5). |
Example:
curl -s "https://proven-decisions.ai/api/v1/content/search?q=shadow+prices&kinds=tutorials,examples&limit=2" \
-H "Authorization: Bearer $PD_KEY"Answer (shortened):
{
"q": "shadow prices",
"kinds": [
"tutorials",
"examples"
],
"counts": {
"tutorials": 2,
"examples": 2
},
"results": [
{
"kind": "tutorials",
"slug": "lp-duality-shadow-prices",
"title": "LP duality and shadow prices",
"snippet": "...l asks: what is one unit of each resource worth? The prices it finds are called shadow prices, and they tell you how much the best profit would rise if you had...",
"level": 4,
"plan": "free",
"url": "https://proven-decisions.ai/api/v1/tutorials/lp-duality-shadow-prices"
},
{
"kind": "tutorials",
"slug": "sensitivity-analysis-what-if",
"title": "Sensitivity analysis: ranging and what-if",
"snippet": "The idea Numbers in a model are estimates: prices move, machines break, suppliers change. Sensitivity analysis asks how much the answer changes when the data c...",
"level": 4,
"plan": "standard",
"url": "https://proven-decisions.ai/api/v1/tutorials/sensitivity-analysis-what-if"
}
]
}GET /api/v1/tutorials
The Learn tutorials or the Refreshers in the website's order, numbered; those above your plan with "locked": true.
| Parameter | Where | What |
|---|---|---|
section | query | learn (default) or refreshers. |
limit | query | Items per page, 1 to 100 (default 20). |
offset | query | How many items to skip (default 0). Use next_offset from the answer for the next page; it is null on the last page. |
Example:
curl -s "https://proven-decisions.ai/api/v1/tutorials?section=refreshers&limit=2" \
-H "Authorization: Bearer $PD_KEY"Answer (shortened):
{
"kind": "refreshers",
"total": 12,
"limit": 2,
"offset": 0,
"next_offset": 2,
"section": "refreshers",
"plan": "standard",
"items": [
{
"slug": "derivative-slope-and-rate",
"number": "1.1",
"title": "The derivative: slope and rate of change",
"section": "refreshers",
"topic": "one variable",
"level": 1,
"plan": "free",
"tags": [
"calculus refresher",
"derivative"
],
"locked": false,
"url": "https://proven-decisions.ai/api/v1/tutorials/derivative-slope-and-rate",
"page_url": "https://proven-decisions.ai/refreshers?t=derivative-slope-and-rate"
},
{
"slug": "critical-points-first-derivative-test",
"number": "1.2",
"title": "Critical points and the first-derivative test",
"section": "refreshers",
"topic": "one variable",
"level": 1,
"plan": "free",
"tags": [
"calculus refresher",
"critical point"
],
"locked": false,
"url": "https://proven-decisions.ai/api/v1/tutorials/critical-points-first-derivative-test",
"page_url": "https://proven-decisions.ai/refreshers?t=critical-points-first-derivative-test"
}
]
}GET /api/v1/tutorials/{slug}
One tutorial or refresher: its text and links to its page and PDF on the site (signed for your plan, valid for 10 minutes).
| Parameter | Where | What |
|---|---|---|
slug | path | From the list or a search. |
format | query | markdown (default; math as LaTeX) or text (plain text, math in Unicode). |
Example:
curl -s "https://proven-decisions.ai/api/v1/tutorials/closed-interval-method" \
-H "Authorization: Bearer $PD_KEY"Answer (shortened):
{
"slug": "closed-interval-method",
"number": "1.6",
"title": "Optimizing on a closed interval: check the end points",
"section": "refreshers",
"topic": "one variable",
"level": 2,
"plan": "free",
"tags": [
"calculus refresher",
"closed interval method"
],
"locked": false,
"url": "https://proven-decisions.ai/api/v1/tutorials/closed-interval-method",
"page_url": "https://proven-decisions.ai/refreshers?t=closed-interval-method",
"markdown": "## The idea\n\nWhen $x$ must stay between two limits, the best point can be where the slope is zero *or* at one of the limits. A continuous function on a closed interval always has a highest and a lowest value, and they can only sit at a few candidates: the critical points inside, and the two end points. Evaluate $f$ at each and compare.\n\n## The math, with a comment on each formula\n\n$$\\min_{a \\le x …",
"html_url": "https://proven-decisions.ai/learn/closed-interval-method.html?p=standard&e=1795015219&s=ff39ecf1e457492d7d60328f2f114a4a",
"pdf_url": "https://proven-decisions.ai/learn/closed-interval-method.pdf?p=standard&e=1795015219&s=ff39ecf1e457492d7d60328f2f114a4a",
"link_expires_at": "2026-11-18T15:20:19+00:00",
"rights": "original, written for OptAI, approved: Derek, 14 Nov 2026"
}A tutorial above your plan, here a Pro tutorial asked for with a Standard key:
{
"detail": "'KKT conditions: how to recognise an optimum' is part of the Pro plan; your plan (Standard) does not include it. See https://proven-decisions.ai/pricing.",
"code": "plan_locked"
}GET /api/v1/examples
The solved examples: classic problems from our golden set, each checked by two solvers.
| Parameter | Where | What |
|---|---|---|
limit | query | Items per page, 1 to 100 (default 20). |
offset | query | How many items to skip (default 0). Use next_offset from the answer for the next page; it is null on the last page. |
Example:
curl -s "https://proven-decisions.ai/api/v1/examples?limit=2" \
-H "Authorization: Bearer $PD_KEY"Answer (shortened):
{
"kind": "examples",
"total": 50,
"limit": 2,
"offset": 0,
"next_offset": 2,
"items": [
{
"slug": "product-mix",
"name": "Product mix problem",
"title": "Product mix problem: a linear programming example, solved and checked",
"class_words": "linear programming",
"problem_class": "LP",
"level": 1,
"discipline": "operations research",
"summary": "Product mix problem worked step by step: the model with a note on every line, the solver's checked answer (HIGHS), why it is optimal, and the dual and code. Change the numbers and …",
"url": "https://proven-decisions.ai/api/v1/examples/product-mix",
"page_url": "https://proven-decisions.ai/examples/product-mix"
},
{
"slug": "diet-problem",
"name": "Diet problem",
"title": "Diet problem: a linear programming example, solved and checked",
"class_words": "linear programming",
"problem_class": "LP",
"level": 1,
"discipline": "operations research",
"summary": "Diet problem worked step by step: the model with a note on every line, the solver's checked answer (HIGHS), why it is optimal, and the dual and code. Change the numbers and solve y …",
"url": "https://proven-decisions.ai/api/v1/examples/diet-problem",
"page_url": "https://proven-decisions.ai/examples/diet-problem"
}
]
}GET /api/v1/examples/{slug}
One solved example: the statement (with units), the model as typed math and LaTeX, the solver's stored result with its proof label, and the walkthrough steps as text. Nothing is solved again; send model_text to POST /api/v1/solve to re-solve it or change the numbers.
| Parameter | Where | What |
|---|---|---|
slug | path | From the list or a search. |
Example:
curl -s "https://proven-decisions.ai/api/v1/examples/product-mix" \
-H "Authorization: Bearer $PD_KEY"Answer (shortened):
{
"slug": "product-mix",
"name": "Product mix problem",
"title": "Product mix problem: a linear programming example, solved and checked",
"class_words": "linear programming",
"problem_class": "LP",
"level": 1,
"discipline": "operations research",
"summary": "Product mix problem worked step by step: the model with a note on every line, the solver's checked answer (HIGHS), why it is optimal, and the dual and code. Change the numbers and …",
"url": "https://proven-decisions.ai/api/v1/examples/product-mix",
"page_url": "https://proven-decisions.ai/examples/product-mix",
"statement": "A company makes products A and B. Each A needs 2 hours of labour and 1 unit of raw material and earns $3 profit; each B needs 1 hour of labour and 2 units of raw material and earns …",
"model_text": "maximize 3*a + 2*b\nsubject to\n2*a + 1*b <= 100\n1*a + 2*b <= 80\na, b >= 0\n# a = number of product A made (units)\n# b = number of product B made (units)\n# objective = total profit ($ …",
"latex": "\\begin{aligned}\\max\\;& 3 a + 2 b\\\\ \\text{s.t.}\\;& 2 a + b \\le 100\\\\ \\;& a + 2 b \\le 80\\\\ \\;& a \\ge 0\\\\ \\;& b \\ge 0\\end{aligned}",
"result": {
"status": "optimal",
"objective": 160.0,
"solution": {
"a": 40.0,
"b": 20.0
},
"solver": "HIGHS",
"problem_class": "LP",
"level": 1,
"verification": {
"status": "optimal",
"max_violation": 0.0,
"integrality_error": 0.0,
"feasible": true,
"n_constraints": 4,
"verified": true,
"dual_check": {
"holds": true,
"gap": 0.0,
"dual_objective": 160.0
}
},
"proof_label": "Exactly proven optimal"
},
"steps": [
{
"title": "Answer",
"text": "Assumed: products can be made in fractional amounts (the optimum is whole anyway). The largest objective value is **160**, at a = 40, b = 20. All 4 constraints are satisfied exactl …"
},
{
"title": "The model",
"text": "Each line has a plain-language note on the right.",
"latex": "\\begin{aligned}\\max\\;& 3 a + 2 b &&\\text{make as large as possible: total profit (\\$)}\\\\ \\text{s.t.}\\;& 2 a + b \\le 100 &&\\text{limit 1: left side cannot exceed 100}\\\\ \\;& a + 2 b …"
}
],
"built_at": "2026-11-10T16:56:30+00:00"
}GET /api/v1/daily
Published Daily Updates, newest first, within the last 365 days.
| Parameter | Where | What |
|---|---|---|
from | query | First day, YYYY-MM-DD. |
to | query | Last day, YYYY-MM-DD. |
limit | query | Items per page, 1 to 100 (default 20). |
offset | query | How many items to skip (default 0). Use next_offset from the answer for the next page; it is null on the last page. |
Example:
curl -s "https://proven-decisions.ai/api/v1/daily?limit=2" \
-H "Authorization: Bearer $PD_KEY"Answer (shortened):
{
"kind": "daily",
"total": 362,
"limit": 2,
"offset": 0,
"next_offset": 2,
"window": {
"from": "2025-11-18",
"to": "2026-11-18"
},
"items": [
{
"slug": "2026-11-17-limited-memory-first-order-methods-a-douglas-rachford-genera",
"title": "Limited-memory first-order methods, a Douglas–Rachford generalisation, and a new Reinhardt proof",
"date": "2026-11-17",
"summary": "A busy arXiv day for first-order theory: limited-memory optimized methods, Douglas–Rachford with unequal step-sizes, and new lower and upper bounds for generalised Lipschitz proble …",
"url": "https://proven-decisions.ai/api/v1/daily/2026-11-17-limited-memory-first-order-methods-a-douglas-rachford-genera",
"page_url": "https://proven-decisions.ai/daily/2026-11-17-limited-memory-first-order-methods-a-douglas-rachford-genera"
},
{
"slug": "2026-11-16-muon-under-shuffling-a-polyak-ojasiewicz-retrospective-and-a",
"title": "Muon under shuffling, a Polyak-Łojasiewicz retrospective, and a polynomial-time result for slope change",
"date": "2026-11-16",
"summary": "A quiet Monday. Theory items include a convergence rate for the Muon optimizer under shuffling, a 10-year look back at the PL condition, and a polynomial-time algorithm for a nonco …",
"url": "https://proven-decisions.ai/api/v1/daily/2026-11-16-muon-under-shuffling-a-polyak-ojasiewicz-retrospective-and-a",
"page_url": "https://proven-decisions.ai/daily/2026-11-16-muon-under-shuffling-a-polyak-ojasiewicz-retrospective-and-a"
}
]
}GET /api/v1/daily/{slug}
One Daily Update as Markdown (any raw HTML from the sources is escaped, as on the website).
| Parameter | Where | What |
|---|---|---|
slug | path | From the list or a search. |
Example:
curl -s "https://proven-decisions.ai/api/v1/daily/2026-11-17-limited-memory-first-order-methods-a-douglas-rachford-genera" \
-H "Authorization: Bearer $PD_KEY"Answer (shortened):
{
"slug": "2026-11-17-limited-memory-first-order-methods-a-douglas-rachford-genera",
"title": "Limited-memory first-order methods, a Douglas–Rachford generalisation, and a new Reinhardt proof",
"date": "2026-11-17",
"summary": "A busy arXiv day for first-order theory: limited-memory optimized methods, Douglas–Rachford with unequal step-sizes, and new lower and upper bounds for generalised Lipschitz proble …",
"url": "https://proven-decisions.ai/api/v1/daily/2026-11-17-limited-memory-first-order-methods-a-douglas-rachford-genera",
"page_url": "https://proven-decisions.ai/daily/2026-11-17-limited-memory-first-order-methods-a-douglas-rachford-genera",
"byline": "OptAI Daily Update",
"checks": "passed",
"markdown": "## The day in optimization\n\nFirst-order theory led the day. Several papers ask how much freedom an algorithm has in its memory, step-sizes or assumptions, and what that freedom buy …"
}GET /api/v1/posts
New in AI: the monthly post on advances of AI in mathematics, newest first.
| Parameter | Where | What |
|---|---|---|
limit | query | Items per page, 1 to 100 (default 20). |
offset | query | How many items to skip (default 0). Use next_offset from the answer for the next page; it is null on the last page. |
Example:
curl -s "https://proven-decisions.ai/api/v1/posts" \
-H "Authorization: Bearer $PD_KEY"Answer (shortened):
{
"kind": "posts",
"total": 1,
"limit": 2,
"offset": 0,
"next_offset": null,
"items": [
{
"slug": "2026-11-02-new-in-ai-openai-s-mass-release-of-machine-written-mathematics-a-quasi",
"title": "OpenAI's mass release of machine-written mathematics: a quasi-Riemann claim and a disputed ETH disproof",
"date": "2026-11-02",
"summary": "OpenAI posted hundreds of manuscripts from an unreleased model, including a quasi-Riemann hypothesis result and a claimed disproof of the Exponential Time Hypothesis. Terence Tao a …",
"url": "https://proven-decisions.ai/api/v1/posts/2026-11-02-new-in-ai-openai-s-mass-release-of-machine-written-mathematics-a-quasi",
"page_url": "https://proven-decisions.ai/ai/2026-11-02-new-in-ai-openai-s-mass-release-of-machine-written-mathematics-a-quasi"
}
]
}GET /api/v1/posts/{slug}
One New in AI post as Markdown.
| Parameter | Where | What |
|---|---|---|
slug | path | From the list or a search. |
Example:
curl -s "https://proven-decisions.ai/api/v1/posts/2026-11-02-new-in-ai-openai-s-mass-release-of-machine-written-mathematics-a-quasi" \
-H "Authorization: Bearer $PD_KEY"Answer (shortened):
{
"slug": "2026-11-02-new-in-ai-openai-s-mass-release-of-machine-written-mathematics-a-quasi",
"title": "OpenAI's mass release of machine-written mathematics: a quasi-Riemann claim and a disputed ETH disproof",
"date": "2026-11-02",
"summary": "OpenAI posted hundreds of manuscripts from an unreleased model, including a quasi-Riemann hypothesis result and a claimed disproof of the Exponential Time Hypothesis. Terence Tao a …",
"url": "https://proven-decisions.ai/api/v1/posts/2026-11-02-new-in-ai-openai-s-mass-release-of-machine-written-mathematics-a-quasi",
"page_url": "https://proven-decisions.ai/ai/2026-11-02-new-in-ai-openai-s-mass-release-of-machine-written-mathematics-a-quasi",
"byline": "OptAI · New in AI",
"checks": "passed",
"markdown": "## The month in AI and mathematics\n\nOctober was dominated by one lab's output. OpenAI says an internal model wrote hundreds of research manuscripts. Many come with Lean proofs, and …"
}GET /api/v1/help
The Help articles, in the Help screen's order.
| Parameter | Where | What |
|---|---|---|
limit | query | Items per page, 1 to 100 (default 20). |
offset | query | How many items to skip (default 0). Use next_offset from the answer for the next page; it is null on the last page. |
Example:
curl -s "https://proven-decisions.ai/api/v1/help?limit=2" \
-H "Authorization: Bearer $PD_KEY"Answer (shortened):
{
"kind": "help",
"total": 10,
"limit": 2,
"offset": 0,
"next_offset": 2,
"items": [
{
"slug": "getting-started",
"title": "Getting started: four ways to enter a problem",
"summary": "ProvenDecisions solves optimization problems: choosing the best option (cheapest, biggest, fastest) under rules you must respect. You can give it a problem in four ways. Pick which …",
"url": "https://proven-decisions.ai/api/v1/help/getting-started",
"page_url": "https://proven-decisions.ai/help/getting-started"
},
{
"slug": "plans-and-limits",
"title": "Plans, prices and limits",
"summary": "Every number here comes from the settings the app enforces. If a number changes, this page changes with it.",
"url": "https://proven-decisions.ai/api/v1/help/plans-and-limits",
"page_url": "https://proven-decisions.ai/help/plans-and-limits"
}
]
}GET /api/v1/help/{slug}
One Help article as Markdown.
| Parameter | Where | What |
|---|---|---|
slug | path | From the list, e.g. plans-and-limits. |
Example:
curl -s "https://proven-decisions.ai/api/v1/help/api-and-mcp" \
-H "Authorization: Bearer $PD_KEY"Answer (shortened):
{
"slug": "api-and-mcp",
"title": "API and MCP server",
"summary": "The ProvenDecisions API and MCP server (Standard and Pro): API keys, solving typed math, reading tutorials, examples and the Daily Update, and the limits.",
"url": "https://proven-decisions.ai/api/v1/help/api-and-mcp",
"page_url": "https://proven-decisions.ai/help/api-and-mcp",
"markdown": "Use ProvenDecisions from your own programs and from AI assistants. The **API** takes a model as typed math or LaTeX and returns the answer as JSON; the **MCP server** gives the sam …"
}GET /api/v1/glossary
The Glossary: every key term and quantity used on the site with its plain-language definition, grouped by topic.
| Parameter | Where | What |
|---|---|---|
limit | query | Items per page, 1 to 100 (default 20). |
offset | query | How many items to skip (default 0). Use next_offset from the answer for the next page; it is null on the last page. |
Example:
curl -s "https://proven-decisions.ai/api/v1/glossary?limit=2" \
-H "Authorization: Bearer $PD_KEY"Answer (shortened):
{
"kind": "glossary",
"total": 153,
"limit": 2,
"offset": 0,
"next_offset": 2,
"items": [
{
"slug": "decision",
"title": "Decision",
"group": "Building a model",
"definition": "A choice you control and want to make well, such as how many loaves to bake, which route to take or how much money to put in each fund. An optimization problem asks for the best se …",
"url": "https://proven-decisions.ai/api/v1/glossary/decision",
"page_url": "https://proven-decisions.ai/glossary?t=decision"
},
{
"slug": "decision-variable",
"title": "Decision variable",
"group": "Building a model",
"definition": "A quantity the solver chooses, standing for one decision; for example x = loaves of bread baked per day. A variable has a unit (loaves, kg, $, hours) and a domain. \"Variable\" on th …",
"url": "https://proven-decisions.ai/api/v1/glossary/decision-variable",
"page_url": "https://proven-decisions.ai/glossary?t=decision-variable"
}
]
}GET /api/v1/glossary/{slug}
One term; the definition as Markdown, with links to related terms on the website.
| Parameter | Where | What |
|---|---|---|
slug | path | From the list, e.g. shadow-price. |
Example:
curl -s "https://proven-decisions.ai/api/v1/glossary/shadow-price" \
-H "Authorization: Bearer $PD_KEY"Answer (shortened):
{
"slug": "shadow-price",
"title": "Shadow price (dual value)",
"group": "Optimality conditions and duality",
"definition": "How much the optimal value changes if a constraint's right-hand side grows by one unit, for example the extra profit from one more oven hour. Valid for small changes.",
"url": "https://proven-decisions.ai/api/v1/glossary/shadow-price",
"page_url": "https://proven-decisions.ai/glossary?t=shadow-price",
"markdown": "How much the optimal value changes if a constraint's [right-hand side](https://proven-decisions.ai/glossary?t=right-hand-side) grows by one unit, for example the extra profit from …"
}GET /api/v1/reports
Update On Demand reports already made (a review of a day, week, month or quarter of the Daily Update). GET /api/v1/reports/{slug} gives one as Markdown.
| Parameter | Where | What |
|---|---|---|
limit | query | Items per page, 1 to 100 (default 20). |
offset | query | How many items to skip (default 0). Use next_offset from the answer for the next page; it is null on the last page. |
Example:
curl -s "https://proven-decisions.ai/api/v1/reports" \
-H "Authorization: Bearer $PD_KEY"Answer (shortened):
{
"kind": "reports",
"total": 3,
"limit": 2,
"offset": 0,
"next_offset": 2,
"items": [
{
"slug": "week-2026-11-09",
"title": "Tighter barriers, sharper rates and a step-size limit for ADMM, while deployment news stays vendor-reported",
"period": "week",
"as_of": "2026-11-09",
"summary": "Theory carried the week: new complexity bounds, convergence results for Newton-type and stochastic methods, and a long-standing ADMM step-size limit revisited. Practice news was th …",
"url": "https://proven-decisions.ai/api/v1/reports/week-2026-11-09",
"page_url": "https://proven-decisions.ai/updates?r=week-2026-11-09"
},
{
"slug": "month-2026-11-09",
"title": "Exact rates, hard limits and GPU solvers: optimization from 10 October to 9 November 2026",
"period": "month",
"as_of": "2026-11-09",
"summary": "The month's theory mostly asked what algorithms cannot do, through matching lower bounds, counterexamples for BFGS, Douglas-Rachford and FTRL, and hardness proofs. GPU-based first- …",
"url": "https://proven-decisions.ai/api/v1/reports/month-2026-11-09",
"page_url": "https://proven-decisions.ai/updates?r=month-2026-11-09"
}
]
}GET /api/v1
The list of endpoints and MCP tools, for programs. No key needed.
No parameters.
Example:
curl -s "https://proven-decisions.ai/api/v1"Answer (shortened):
See /api/v1.
MCP tools
The MCP server at https://proven-decisions.ai/mcp (streamable HTTP, the same key in the Authorization header, or an OAuth sign-in for claude.ai custom connectors) gives AI assistants the same functions; setup for Claude Code, Claude Desktop and claude.ai is on the Developers page. Tool calls count against the same limits as the REST calls (solving tools against the solve limits, content tools against the lighter limit); listing the tools is free.
| Tool | Arguments | Returns |
|---|---|---|
solve | model, tier? | status, objective, solution, proof_label, verification, problem_class, level, solver |
dual | model, tier? | The dual, solved, with the duality report |
kkt | model, tier? | Multipliers, residual, holds |
explain | model, depth? (1-10, default 3), tier? | The walkthrough as Markdown, with status, objective and proof label |
search_content | q, kinds? (list) | Up to 5 hits per kind, as GET /api/v1/content/search |
get_content | kind, slug | One item, as the matching GET endpoint |
list_content | kind, section?, limit?, offset? | One page of a list, as the matching GET endpoint |
Example: asked “What do the tutorials say about shadow prices?”, an assistant calls
{
"name": "search_content",
"arguments": {
"q": "shadow prices",
"kinds": [
"tutorials"
]
}
}then reads the best hit with
{
"name": "get_content",
"arguments": {
"kind": "tutorials",
"slug": "lp-duality-shadow-prices"
}
}Each answer is the same JSON as the REST endpoint, as text. A refusal (no key, plan, locked tutorial, limit) comes back as a tool error with the same message.
OpenAPI document
https://proven-decisions.ai/api/v1/openapi.json describes the v1 endpoints (parameters, answers and errors) in OpenAPI 3; import it into Postman, Insomnia or a code generator. No key is needed to read it.
More: Developers · API and MCP in Help · Plans and limits · Pricing
Try ProvenDecisions free: type, photograph or say an optimization problem; every answer is checked by a solver.