Aturan.org – Indonesian Legal Research
Large-scale semantic search and article-level retrieval for Indonesian regulations, providing AI agents with structured access to relevant laws and authoritative legal text.
Hosted MCP Server
npx add-mcp 'https://mcp.aturan.org/mcp'Installs into Claude Code, Codex, Cursor and more
Documentation
Aturan.org
Semantic Legal Retrieval for Indonesian Regulations
Find legal norms by meaning. Verify the source.
Grounded regulatory data for people, applications, and AI agents.
Website · MCP Server · REST API · Discover
What is Aturan.org?
Aturan.org is a semantic legal retrieval system for Indonesian regulations.
It helps people, applications, and AI agents discover regulations and legal provisions based on meaning and legal context, rather than relying only on exact keywords or already knowing the title of a regulation.
Aturan.org helps match legal issues with candidate provisions. Users can identify the relevant regulations and read the complete provisions before using them in legal analysis.
The platform is available through:
- a public web application for legal research;
- a REST API for application and system integration; and
- a remote Model Context Protocol (MCP) server for AI clients and agentic workflows.
Aturan.org is a retrieval system, not an automated legal-opinion service.
Semantic retrieval identifies relevant legal material. Legal applicability and interpretation still require evaluation of the complete provision, regulatory context, amendments, hierarchy, and the facts being analysed.
Legal Knowledge Coverage
Aturan.org provides semantic retrieval across a continuously updated corpus of Indonesian regulations and legal provisions, covering national, regional, and institutional regulatory frameworks.
Corpus Overview
| Metric | Coverage |
|---|---|
| Indonesian regulations | 296,433+ |
| Legal provisions (Pasal) | 5,684,209+ |
| Data updates | Daily |
| Vector database size | 73.9 GB+ |
Regulatory Coverage
The corpus spans national legislation, presidential and ministerial instruments, regional regulations, and regulations issued by state institutions and public authorities.
The following table highlights major regulatory categories with published coverage statistics.
| Regulation Type | Regulations | Pasal |
|---|---|---|
| Laws (UU) | 1,924 | 50,431 |
| Government Regulations in Lieu of Law (PERPU) | 164 | 2,923 |
| Government Regulations (PP) | 4,988 | 90,720 |
| Presidential Regulations (PERPRES) | 2,654 | 43,539 |
| Presidential Decrees (KEPPRES) | 6,823 | 19,988 |
| Presidential Instructions (INPRES) | 461 | 116 |
| Ministerial Regulations (PERMEN) | 17,324 | 471,265 |
| Ministerial Decrees (KEPMEN) | 397 | 1,059 |
| Regional Regulations (PERDA) | 61,688 | 1,765,861 |
| Governor Regulations (PERGUB) | 18,370 | 319,086 |
| Mayor Regulations (PERWALI) | 35,882 | 554,687 |
| Regent Regulations (PERBUP) | 138,839 | 2,193,172 |
Additional Regulatory Instruments
Beyond the major categories listed above, Aturan.org also covers a broader range of legal and administrative instruments, including:
- Special regional legislation: Qanun, Papua Special Regional Regulations (Perdasus), and Special Region Regulations.
- Judicial institutions: Regulations issued by the Supreme Court and other judicial authorities.
- Law enforcement institutions: Regulations and decisions issued by the Attorney General's Office and National Police.
- State institutions and commissions: Regulations issued by constitutional bodies, commissions, councils, and independent state institutions.
- Government agencies and public authorities: Regulations and decisions issued by agencies, boards, supervisory authorities, and other public institutions.
- Other administrative instruments: Circular letters, instructions, decrees, and institutional regulations.
The coverage statistics above represent selected major categories and do not constitute an exhaustive breakdown of the entire regulatory corpus.
Semantic Legal Discovery
Aturan.org helps AI agents and legal researchers discover relevant legal provisions across national, regional, and institutional regulations without knowing the regulation title or exact legal wording.
The regulatory corpus is continuously maintained and expanded through daily synchronization.
Why semantic legal retrieval?
Legal research does not always begin with the name of a regulation.
It often begins with an issue:
- What obligations apply?
- What activity is prohibited?
- What authority does an institution have?
- What procedure must be followed?
- What sanctions may apply?
- What regulations govern a particular activity?
Traditional title or keyword search works well when the researcher already knows what document or wording to search for.
But legal questions are often expressed differently from the language used by legislation.
Aturan.org addresses this by matching the meaning of a query against structured legal provisions.
Legal question or concept
│
▼
Semantic retrieval
│
▼
Candidate Pasal
│
▼
Relevant regulations
│
▼
Read full provisions
│
▼
Analyze
This makes it possible to begin legal research from the substance of the issue, rather than from an assumed document.
What you can do
| Capability | Purpose |
|---|---|
| Regulation discovery | Discover regulations connected to a legal issue across regulation types and hierarchies. |
| Semantic provision discovery | Find candidate provisions related to a right, obligation, prohibition, authority, procedure, sanction, condition, or other legal concept. |
| Regulation identity search | Locate a regulation when its type, number, year, title, or title fragment is already known. |
| Full-Pasal retrieval | Read the complete wording of a selected provision before quoting or analysing it. |
| Legal landscape research | Explore an issue across multiple regulations and normative dimensions. |
| AI grounding | Give AI agents structured access to retrieved Indonesian legal material before reasoning. |
| Application integration | Integrate Indonesian regulatory retrieval into LegalTech, RegTech, RAG, research, and other applications. |
| Research workflow | Search, inspect, save, export, and continue working with retrieved legal material. |
Built for people, applications, and AI
Aturan.org exposes the same legal-retrieval foundation through three different interfaces.
Web
The web application is designed for people conducting legal and regulatory research.
Users can search by legal issue, inspect regulations and Pasal candidates, verify source material, save relevant provisions, and export research results.
Website:
https://aturan.org
REST API
The REST API provides programmatic access to Aturan.org retrieval services.
It is intended for applications, LegalTech, RegTech, RAG pipelines, enterprise systems, and developers who need explicit control over requests and responses.
Documentation:
https://aturan.org/api
MCP Server
The Aturan.org MCP Server exposes legal retrieval as tools that AI models and agents can call directly.
Instead of placing an entire legal corpus into a model context, an agent can retrieve only the regulatory material needed for the current question, inspect the results, read selected provisions, and continue its reasoning from grounded source material.
Documentation:
https://aturan.org/mcp
Connect your AI
Aturan.org provides a remote MCP server using Streamable HTTP.
https://mcp.aturan.org/mcp
Quick install
Connect Aturan.org to supported AI development environments using a single command:
npx add-mcp 'https://mcp.aturan.org/mcp'
The add-mcp CLI simplifies MCP server configuration for compatible AI clients and development tools.
After installation, complete authentication using OAuth or an API Key, depending on the client's capabilities.
Alternatively, configure the MCP endpoint manually in any compatible client supporting remote MCP over Streamable HTTP.
For client-specific instructions, see the MCP documentation.
Authentication
Aturan.org supports two authentication paths:
Aturan.org Account
│
┌──────────────┴──────────────┐
│ │
OAuth API Key
│ │
interactive access programmatic access
│ │
└──────────────┬──────────────┘
│
▼
Aturan.org services
OAuth
OAuth provides an interactive connection between a compatible client and an Aturan.org account.
The client can open the Aturan.org authorization flow in a browser, where the user signs in and grants access. The client can then manage the resulting OAuth credentials according to its own implementation.
OAuth is particularly convenient for AI services and MCP clients with native OAuth support.
API Key
Users can create dedicated API keys directly from the Aturan.org Dashboard.
API keys are suitable for:
- MCP clients using bearer authentication;
- REST API integration;
- backend applications;
- automation and agent services;
- development environments; and
- server-to-server workflows.
Separate keys can be created for different clients or environments and revoked independently when no longer needed.
API keys are transmitted using the HTTP Authorization header:
Authorization: Bearer YOUR_API_KEY
Do not embed API keys in public source code or expose them in client-side applications.
For current authentication details, supported clients, connection instructions, and configuration examples, see:
For REST API integration, see:
Operational connection details may evolve independently from this repository. The MCP and REST API documentation are the source of truth for their respective interfaces.
MCP tools
The MCP server exposes five complementary legal retrieval tools.
| Tool | Purpose |
|---|---|
cari_peraturan_terkait | Discover regulations related to a legal issue through semantic search. |
cari_pasal_terkait | Find candidate Pasal directly by semantic similarity to a legal concept or normative issue. |
cari_judul_peraturan | Resolve a known regulation from its title, type, number, year, or title fragment. |
baca_isi_pasal | Retrieve the complete wording of one selected Pasal from a regulation already identified by Aturan.org. |
baca_isi_pasal_batch | Retrieve multiple already-identified Pasal in one call, including across different regulations. |
Each tool answers a different retrieval question.
Which regulations may matter?
│
└── cari_peraturan_terkait
Which norms may matter?
│
└── cari_pasal_terkait
Which exact regulation is this?
│
└── cari_judul_peraturan
What does one provision actually say?
│
└── baca_isi_pasal
What do several identified provisions say?
│
└── baca_isi_pasal_batch
The tools are complementary rather than interchangeable. An agent does not need to call every tool for every question. It should select the tools that add information needed for the research task.
When several relevant Pasal have already been identified, baca_isi_pasal_batch can retrieve up to 20 provisions in a single call, including provisions from different regulations.
Retrieval workflow
A useful mental model for legal research with Aturan.org is:
DISCOVER → IDENTIFY → READ → ANALYZE
1. Discover
Find candidate regulations or provisions.
Depending on what is already known, this may begin with:
cari_peraturan_terkait
cari_pasal_terkait
cari_judul_peraturan
2. Identify
Evaluate the retrieved candidates.
Relevant signals may include:
- regulation title;
- regulation type;
- year;
- status;
- metadata;
- semantic hits;
- candidate Pasal numbers; and
- relationships with other retrieved regulations.
3. Read
Retrieve the complete wording of provisions that will support the analysis.
For a single provision:
baca_isi_pasal
When several relevant Pasal have already been identified:
baca_isi_pasal_batch
The batch tool can read up to 20 identified Pasal in one call, including Pasal from different regulations.
A semantic hit is a candidate, not a substitute for reading the legal provision itself.
4. Analyze
Only after retrieving the relevant material should the model or researcher build the substantive analysis.
The analysis should distinguish between:
retrieved legal text
↓
relationship between regulations
↓
interpretation
↓
analytical conclusion
Three retrieval paths
The appropriate workflow depends on what is already known.
Known regulation
If the regulation is already identified:
Known regulation
↓
cari_judul_peraturan
↓
regulation_id
↓
baca_isi_pasal
↓
analyze
Known legal issue
If the issue or norm is known but the regulation is not:
Known legal issue
↓
cari_pasal_terkait
↓
evaluate candidates
↓
baca_isi_pasal
or baca_isi_pasal_batch
↓
analyze
Regulatory landscape
If the objective is to identify the broader regulatory framework:
Regulatory landscape
↓
cari_peraturan_terkait
↓
evaluate regulation groups
↓
inspect candidate Pasal
↓
baca_isi_pasal_batch
when multiple Pasal are identified
↓
analyze
Complex legal issues can combine these paths and run several focused retrievals before analysis.
Semantic retrieval
Aturan.org semantic search is designed around the substance of legal norms.
A useful query describes the legal concept being sought.
For example:
hak ahli waris penerima manfaat program jaminan kematian BPJS Ketenagakerjaan
rather than:
bagaimana jaminan kematian BPJS Ketenagakerjaan untuk ahli waris?
Or:
kewajiban pemberi kerja mendaftarkan pekerja dalam program jaminan sosial
rather than:
apa kewajiban perusahaan soal BPJS?
Semantic retrieval benefits from sufficient context. Queries for semantic retrieval should therefore contain enough substantive information to represent the intended legal meaning.
For complex issues, several focused queries are usually more useful than one oversized query containing every aspect of the problem.
Current query requirements and retrieval parameters are documented in the relevant API and MCP specifications.
Multilingual semantic retrieval
Semantic retrieval is based on meaning rather than literal word matching.
This allows queries to be expressed differently from the wording found in Indonesian regulations, including queries written in other languages supported by the underlying multilingual embedding model.
For example:
foreign investor land ownership restrictions
can be used as a semantic query against Indonesian regulatory provisions even though the source legislation itself is written in Indonesian.
English or multilingual query
│
▼
semantic embedding
│
▼
Indonesian legal corpus
│
▼
candidate Pasal
This is useful for international users, cross-border research, and AI agents that may reason in a language different from the language of the underlying legal source.
Semantic similarity, however, is not legal interpretation. Retrieved candidates must still be evaluated in their complete regulatory context.
Agentic legal research
Aturan.org is designed for multi-step retrieval, not only one-shot search.
An AI agent can use its own reasoning to formulate queries, inspect retrieval results, identify missing dimensions, retrieve again, and read selected provisions before producing an answer.
For example, a question about rooftop solar regulation may lead an agent to investigate several dimensions:
pengaturan PLTS atap
perizinan PLTS atap
kewajiban pemegang izin usaha penyediaan tenaga listrik terkait PLTS atap
ekspor impor energi listrik PLTS atap
kapasitas pemasangan PLTS atap
The agent can then compare the regulations and Pasal candidates returned from those retrievals and read the provisions that materially support the analysis.
The workflow becomes:
question
│
▼
understand the issue
│
▼
formulate focused queries
│
▼
retrieve
│
▼
inspect results
│
├──────── insufficient coverage ────────┐
│ │
▼ │
refine query │
│ │
└──────────── retrieve again ◄──────────┘
│
▼
read selected provisions
│
▼
compare legal material
│
▼
analyze
This allows the model's reasoning capability and Aturan.org's retrieval capability to perform separate jobs:
The model reasons about what to investigate; Aturan.org retrieves the regulatory evidence.
Principles for AI agents
Retrieve before reasoning
When an analysis depends on Indonesian regulations, retrieve the relevant legal material before relying on the model's internal knowledge.
Internal model knowledge can help formulate queries, decompose an issue, recognize legal concepts, and reason about the retrieved material.
It should not replace retrieval when a regulation, Pasal, or legal wording is being presented as the basis of the analysis.
Read before citing
Semantic search identifies candidates.
Before quoting a provision or relying on its wording, retrieve the complete Pasal.
semantic hit ≠ full legal provision
Do not fabricate references
Regulation identities, Pasal numbers, regulation_id values, and legal text presented as Aturan.org retrieval results must come from actual retrieval results.
In particular:
Never guess regulation_id.
Iterate when necessary
Legal retrieval is not necessarily a one-query process.
query
↓
inspect
↓
refine
↓
retrieve again
↓
read
↓
analyze
The number of retrievals should depend on the complexity of the issue and whether each additional call contributes useful evidence.
Retrieval capabilities
Aturan.org supports complementary capabilities for discovering regulations, identifying relevant provisions, and reading source material.
| Capability | Purpose |
|---|---|
| Regulation identity | Resolve a known regulation from its title, type, number, year, or title phrase. |
| Norm discovery | Find candidate provisions relevant to a legal concept. |
| Landscape discovery | Discover regulations connected to a legal issue. |
| Source verification | Read complete provisions before citation or analysis. |
These capabilities are available through the Web application, REST API, and MCP Server.
Indonesian regulatory corpus
│
▼
Aturan.org
│
┌──────────┼──────────┐
▼ ▼ ▼
Web REST API MCP Server
│ │ │
▼ ▼ ▼
Researchers Applications AI clients
& agents
GPU-accelerated exact search
Aturan.org uses GPU-accelerated Exact Nearest Neighbor (ENN) search rather than approximate nearest-neighbor indexing. This reflects our commitment to thorough semantic candidate discovery for legal research.
query
│
▼
query embedding
│
▼
GPU-accelerated exact search
│
▼
nearest candidates
GPU acceleration makes exhaustive similarity computation practical at corpus scale, supporting responsive legal retrieval while preserving exact nearest-neighbor search.
Discovery and verification are separate
Semantic retrieval answers:
What legal material should I inspect?
Full-Pasal retrieval answers:
What does the selected provision actually say?
Keeping those operations separate prevents a similarity result or snippet from being treated as though it were the complete legal rule.
Designed for efficient agentic use
Aturan.org does not assume that legal retrieval requires the largest possible language model.
The MCP workflow is deliberately structured so that retrieval tools have clear roles, bounded inputs, and machine-readable outputs.
This makes the system suitable for tool-calling agents ranging from lightweight local models to larger frontier models.
A capable agent needs to understand a relatively small set of operations:
discover regulations
find norms
resolve regulation identity
read provisions
The legal corpus and retrieval workload remain outside the LLM itself.
This separation allows the language model to concentrate on planning and reasoning while Aturan.org performs legal evidence retrieval.
Institution and enterprise use
The retrieval architecture is designed so that the same general pattern can be used beyond the public Aturan.org service.
structured legal corpus
+
semantic retrieval
+
legal source reading
+
tool interface
+
agentic LLM
This architecture is suitable for LegalTech, RegTech, institutional legal knowledge systems, private RAG environments, and on-premise AI deployments where organizations require greater control over infrastructure, governance, or data boundaries.
The public Aturan.org service demonstrates this retrieval model using Indonesian regulations.
Interfaces
| Interface | Address | Primary role |
|---|---|---|
| Web | https://aturan.org | Human-facing semantic legal research. |
| MCP Server | https://aturan.org/mcp | AI clients and agentic legal retrieval. |
| REST API | https://aturan.org/api | Application and system integration. |
| Discover | https://aturan.org/discover | Overview of available integration methods and use cases. |
The interfaces serve different users but share the same core principle:
find relevant legal material
↓
inspect the source
↓
read the norm
↓
then analyze
Source verification
Aturan.org is designed to support grounded legal research, not to replace authoritative legal sources.
Where source verification is required, users and AI systems should inspect the official source document associated with the retrieved regulation.
A retrieval result helps locate relevant legal material.
It does not transform Aturan.org into the official publisher of that material.
Limitations
Semantic similarity is evidence of retrieval relevance, not evidence of legal applicability.
A highly similar provision does not by itself establish that the provision:
- applies to every factual situation;
- is the primary or controlling legal basis;
- remains effective without amendment;
- overrides another regulation;
- has no relevant implementing regulation;
- supports a particular interpretation; or
- leads to a particular legal conclusion.
Before drawing a legal conclusion, evaluate:
- the complete wording of the Pasal;
- the context of the provision within the regulation;
- definitions and related provisions;
- the regulation's status;
- amendments and revocations;
- implementing regulations;
- relationships between regulations;
- regulatory hierarchy; and
- the facts being analysed.
Aturan.org helps retrieve the evidence needed for that work.
The interpretation remains a separate analytical step.
Documentation
Detailed and current specifications are maintained outside this README.
MCP Server
Connection methods, OAuth, API Key authentication, MCP tools, parameters, retrieval strategy, and AI usage guidance:
REST API
Authentication, endpoints, request parameters, response structures, and integration guidance:
Discover
Comparison of the available interfaces and guidance on choosing between Web, REST API, and MCP:
These documentation pages are the source of truth for operational specifications that may change over time.
Service access
Aturan.org provides free and paid access to its hosted services.
Access conditions, authentication methods, credits, quotas, usage limits, and available plans may differ between the Web application, REST API, and MCP Server.
Users can manage their account and create or revoke API keys through the Aturan.org Dashboard.
Refer to the relevant Aturan.org documentation for current service specifications.
About this repository
This repository represents the public-facing Aturan.org project and its published resources.
The hosted Aturan.org service includes infrastructure, datasets, retrieval systems, and backend components that are not necessarily part of this repository.
The Aturan.org name and logo are proprietary brand assets.
Any repository license applies only to source code and documentation explicitly published under that license. It does not grant rights to private backend systems, hosted services, datasets, trademarks, or brand assets unless expressly stated otherwise.
Find the norm by meaning. Verify the source.