--- title: "Atlan vs TrustGraph: Context Layer vs Self-Hosted Graph [2026]" url: "https://atlan.com/know/ai-agent/knowledge-graph/atlan-vs-trustgraph/" description: "Atlan vs TrustGraph compared: what TrustGraph's open-source context graph does, what Atlan's context layer adds, and how both work together in 2026." author: "Ayswarrya G" author_role: "Contributing Writer, Data Engineering & Metadata" published: "2026-10-12" updated: "2026-10-12" --- > Atlan is hosting Context Conference, bringing together the leaders and builders at the frontier of giving AI the context it needs to understand their business. It runs online on October 28, 2026, from 11:00 AM to 2:00 PM ET. Atlan co-founder Prukalpa Sankar opens and closes the day. Leaders from AstraZeneca, BNY and Verizon share why they invest in context and what they get from it. Registrants get early access to The AI Context Gap, a new study from MIT Technology Review Insights. Register: https://atlan.com/context-conference/ --- Two products now use the same vocabulary for different jobs. TrustGraph is the [context graph](https://atlan.com/know/what-is-a-context-graph/) a platform team assembles from its documents, and Atlan is the [Context Layer for AI](https://atlan.com/know/what-is-context-layer/) a data organization runs across its warehouses, pipelines, and BI tools. The useful question is which questions your agents need answered, and where the definitions behind those answers live. ### Score Whether an Agent Can Answer Your Question Give it one business question, such as a loss ratio, and the artifacts you can produce. It returns a score per dimension, the evidence each one needs, and the work list. [Read the skill](/skills/ai-ready-question-score.md). *Paste into a new chat* ``` Use the skill at https://atlan.com/skills/ai-ready-question-score.md to score our data estate for one question. Ask me for whatever it needs. ``` *Run once in a terminal* ``` curl -fsSL --create-dirs \ -o ~/.agents/skills/ai-ready-question-score/SKILL.md \ https://atlan.com/skills/ai-ready-question-score.md ``` *For an agent* ``` curl -fsSL https://atlan.com/skills/ai-ready-question-score.md ``` Human Agent ClaudeChatGPTGeminiCursorOther Copy .skp{margin:1.75rem 0;padding:1rem 1.125rem 1.125rem} .skp h3{margin:0 0 .375rem!important;font-size:20px!important;line-height:24px!important; font-family:var(--font-funnel-display)!important;font-weight:600!important;color:#2B2B39!important} .skp > p{margin:0 0 .875rem!important;font-size:14px!important;line-height:22px!important;color:#555572!important} .skp > p a{color:#2026D2;text-decoration:underline} .skp-box{margin-bottom:.75rem} .skp-p em{display:block;font-size:10px;font-weight:700;letter-spacing:.06em;text-transform:uppercase; color:#F34D77;font-style:normal;margin-bottom:.25rem} .skp-p > p{margin:0!important} #seo-layout article .skp .skp-p pre{margin:0!important;overflow-x:auto} #seo-layout article .skp .skp-p pre code{color:#2B2B39!important;font-size:12px!important;line-height:19px; white-space:pre-wrap;overflow-wrap:anywhere;background:none;padding:0; font-family:ui-monospace,SFMono-Regular,Menlo,monospace} .skp-p + .skp-p{margin-top:.625rem} .skp-cs{display:flex;flex-wrap:wrap;align-items:center;gap:.375rem} .skp-seg{display:inline-flex;flex:0 0 auto;border:1px solid #DDDDE3;border-radius:999px;overflow:hidden;background:#fff} .skp-m{min-height:32px;padding:.25rem .75rem;border:0;background:transparent;font-family:inherit;color:#77778E;cursor:pointer} .skp-m:hover{color:#2026D2} .skp-m.on{background:#2026D2;color:#fff} .skp-sep{flex:0 0 auto;width:1px;height:20px;background:#DDDDE3;margin:0 .25rem} .skp-tools{display:inline-flex;flex-wrap:wrap;gap:.375rem;min-width:0} .skp-cp{margin-left:auto;min-height:32px;padding:.25rem .75rem;border-radius:999px;background:#fff; font-family:inherit;color:#555572;cursor:pointer} .skp-cp:hover{border-color:#2026D2!important;color:#2026D2} .skp-cs .skp-c{display:inline-flex!important;width:auto;flex:0 0 auto;align-items:center;gap:.375rem; min-height:32px;padding:.25rem .625rem;border-radius:999px;background:#fff;font-family:inherit; color:#555572;cursor:pointer} .skp-cs .skp-c img{display:block!important;width:14px!important;height:14px!important; max-width:14px!important;flex:0 0 auto;opacity:.55;margin:0} .skp-c:hover img,.skp-c.on img{opacity:1} .skp-cs .skp-c:hover{border-color:#797DE4!important;color:#2026D2} .skp-cs .skp-c.on{border-color:#2026D2!important;background:#F4F4FD;color:#2026D2} .skp-agent .skp-tools,.skp-agent .skp-sep{display:none} (function(){ var p=document.querySelector('.skp[data-skill="ai-ready-question-score"]');if(!p)return; var panes=p.querySelectorAll('.skp-p'); function show(id){panes.forEach(function(x){x.hidden=x.dataset.skp!==id;});} function cur(){var a=p.querySelector('.skp-p:not([hidden]) pre code');return a?a.textContent:'';} show('prompt'); p.querySelectorAll('.skp-p em').forEach(function(e){ var par=e.parentElement; ((par.tagName==='P'&&par.textContent.trim()===e.textContent.trim())?par:e).hidden=true; }); p.querySelectorAll('.skp-c').forEach(function(b){ b.addEventListener('click',function(){ p.querySelectorAll('.skp-c').forEach(function(x){var on=x===b;x.classList.toggle('on',on);x.setAttribute('aria-pressed',on);}); show(b.dataset.skp); }); }); p.querySelectorAll('.skp-m').forEach(function(m){ m.addEventListener('click',function(){ var agent=m.dataset.skpMode==='agent'; p.classList.toggle('skp-agent',agent); p.querySelectorAll('.skp-m').forEach(function(x){var on=x===m;x.classList.toggle('on',on);x.setAttribute('aria-pressed',on);}); if(agent){show('agent');}else{var s=p.querySelector('.skp-c.on');show(s?s.dataset.skp:'prompt');} }); }); var c=p.querySelector('[data-skp-copy]'); c.addEventListener('click',function(){ navigator.clipboard.writeText(cur()).then(function(){ c.textContent='Copied';setTimeout(function(){c.textContent='Copy';},1600); }); }); })(); --- ## What does TrustGraph do? TrustGraph is an open-source ([Apache 2.0](https://github.com/trustgraph-ai/trustgraph)) [context orchestration layer](https://docs.trustgraph.ai/learn/how-it-works). It uses LLMs to extract documents and data files into an ontology-grounded hypergraph, then serves that knowledge to agents through [Graph RAG, Ontology RAG, or Document RAG](https://docs.trustgraph.ai/learn/how-it-works). Graph-based context improves accuracy because the model gets explicit relationships between facts instead of loose text chunks. Microsoft Research found that [baseline RAG "struggles to connect the dots"](https://www.microsoft.com/en-us/research/blog/graphrag-unlocking-llm-discovery-on-narrative-private-data/) when an answer depends on linking scattered information. TrustGraph is built for that gap. It extracts entities and relationships into a graph, gives agents a connected subgraph of facts for each question, and its [explainability documentation](https://docs.trustgraph.ai/architecture/explainability) describes tracing any edge back through the chunk, page, and document it came from. For the retrieval side of the story, see [what GraphRAG is](https://atlan.com/know/what-is-graphrag/) and how [a context layer differs from RAG](https://atlan.com/know/ai-agent/agent-context-layer-vs-rag/). What Is a Context Graph? The AI Reasoning Layer Explained ### What TrustGraph leaves to the team that runs it * **Warehouse and BI coverage**: Snowflake, Databricks, dbt, Tableau, Power BI, and Looker do not appear on its [integrations page](https://trustgraph.ai/integrations/). Extending TrustGraph means writing a [custom processor](https://docs.trustgraph.ai/guides/building/custom-processing.html), which is a Python class. * **Structured data at scale**: Structured loading works from CSV, JSON, and XML files, and the [load guide](https://docs.trustgraph.ai/guides/structured-processing/load-file.html) says the schema has to be created by hand first. That guide was written against an earlier release, so check the current docs for your version. * **Ontology authoring**: Teams [bring their own OWL ontology](https://docs.trustgraph.ai/learn/ontologies.html). The docs note that a code assistant can draft one from a description of the domain, so the work shifts to reviewing it and keeping it current as the business changes. * **Data lineage**: [Provenance](https://docs.trustgraph.ai/architecture/explainability) traces an answer back to a source document and chunk. The documentation does not describe tracing a metric through tables, transformations, and dashboards. * **Running the stack**: A self-hosted deployment means the owning team sizes, patches, and monitors [Cassandra, Qdrant, Garage, and the Pulsar or RabbitMQ messaging layer](https://docs.trustgraph.ai/architecture/storage.html), plus the inference services. * **Upgrade work**: Releases can break integrations. The [2.9 release notes](https://trustgraph.ai/news/release-2-9) list removed LangChain dependencies, a dropped `trustgraph-unstructured` package, and renamed Workbench component IDs, so custom workflows need rework at each upgrade. --- ## What does Atlan's context layer add on top of TrustGraph? Atlan works one level above a graph that a single team assembles. It builds context from the systems the business already runs, puts domain owners in charge of certifying it, and serves the same answer to every agent. For a team already running TrustGraph, the additions fall into four products and one delivery path. ### Enterprise Data Graph and column-level lineage Native connectors cover warehouses (Snowflake, BigQuery, Redshift, Databricks), BI tools (Tableau, Power BI, Looker, Sigma), pipelines (dbt, Fivetran, Matillion), and orchestrators (Airflow, Dagster), with [100+ in total](https://atlan.com/connectors/). The [Enterprise Data Graph](https://atlan.com/know/enterprise-data-graph/) sits on top of them. [Column-level lineage](https://atlan.com/data-lineage/) is built automatically from SQL, pipelines, and BI tools, which is the half of the estate a document-extraction pipeline does not see. See [why agents need it](https://atlan.com/know/ai-agent/data-for-ai/data-lineage-for-ai/). Niche or homegrown systems get custom connectors through the open-source [Application SDK](https://github.com/atlanhq/application-sdk). ### Context Agents [Nine specialist agents](https://atlan.com/context-agents/) (Scout, Scribe, Lexis, Doc, Nexus, Sage, Atlas, Vera, and Orion) read query history, SQL, lineage, and dbt logic to write descriptions, build a glossary, resolve metric conflicts, tag domains, score quality, and map ontology. Atlan reports 350K+ AI descriptions applied at a 90%+ acceptance rate, and 87% of customers say the agents write higher quality content than humans, according to the [Context Agents page](https://atlan.com/context-agents/). The OWL ontology TrustGraph asks a team to author is bootstrapped here from signals the estate already holds. For the concept, see [what an ontology does for AI](https://atlan.com/know/what-is-ontology-in-ai/). ### Context Engineering Studio This is the [context lifecycle layer](https://atlan.com/context-engineering-studio/): bootstrap, simulate, deploy, observe. [Context Repos](https://atlan.com/know/ai-agent/context-repository-for-ai-agents/) are versioned, domain-scoped units with Git-like branching, staged rollouts, and rollbacks, and domain experts approve changes inside Bounded Context Spaces. The studio uses existing dashboards and reports to simulate test questions, and when a user marks an agent answer wrong, it generates a suggested context update. ### Context Lakehouse A knowledge graph, Apache Iceberg storage, and vector-native search [sit in one store](https://atlan.com/context-lakehouse/), reachable over MCP, A2A, SQL, and REST or Graph APIs. Iceberg time travel preserves every version of every asset state, which supports GDPR deletion audits, CCPA access trails, and SOX change logs. Because the files are open Iceberg tables, metadata lives in your own cloud storage and stays readable by any Iceberg-compatible engine. ### Context delivery [Atlan MCP](https://docs.atlan.com/product/capabilities/atlan-ai/how-tos/atlan-mcp-overview) is a hosted server on every tenant, so Claude, Cursor, ChatGPT, Gemini, and Copilot connect without new infrastructure, and every action respects the permissions already set in Atlan. Our [guide to Atlan MCP](https://atlan.com/know/what-is-atlan-mcp/) covers setup. MCP itself became vendor-neutral when Anthropic [donated it to the Linux Foundation's Agentic AI Foundation](https://www.anthropic.com/news/donating-the-model-context-protocol-and-establishing-of-the-agentic-ai-foundation) in December 2025, which keeps any context source readable by a [model-agnostic](https://atlan.com/know/ai-agent/context-layer/model-agnostic-context-layer/) agent stack. --- ## Atlan vs TrustGraph: how do the key capabilities compare? Both products are evolving quickly, so this table maps each dimension of TrustGraph against the Atlan context layer. Where TrustGraph's documentation is silent on a point, the cell says so rather than guessing. | Capability | TrustGraph | Atlan context layer | Notes | |---|---|---|---| | **Positioning** | ["Open-source context orchestration layer"](https://docs.trustgraph.ai/learn/how-it-works) (docs); ["Semantic Intelligence Layer for Ontologies"](https://trustgraph.ai/) (homepage) | [Context Layer for AI](https://atlan.com/context-layer/) | Shared vocabulary, different scope | | **License and hosting** | [Apache 2.0](https://github.com/trustgraph-ai/trustgraph); self-hosted with [Docker, Kubernetes, or the cloud providers its docs list](https://docs.trustgraph.ai/deployment/) | Managed platform; [Context Lakehouse](https://atlan.com/context-lakehouse/) in open Iceberg format, in your own cloud storage | TrustGraph suits deployments your team operates end to end | | **Where context comes from** | Documents and files via LLM extraction; CSV, JSON, and XML via [defined schemas](https://docs.trustgraph.ai/guides/structured-processing/load-file.html) | [100+ connectors](https://atlan.com/connectors/); lineage, SQL, BI semantics, and usage | The deciding row for most buyers | | **Warehouse and BI connectivity** | Not on the [integrations page](https://trustgraph.ai/integrations/); extended through custom processors | Native: Snowflake, BigQuery, Databricks, dbt, Tableau, Power BI, Looker, and more | Warehouse and BI hold most metric definitions | | **Ontology** | [Bring your own OWL ontology](https://docs.trustgraph.ai/learn/ontologies.html); Workbench and CLI tooling to load and manage it | [Orion and Lexis agents](https://atlan.com/context-agents/) bootstrap ontology and glossary from estate signals | Hand-authored versus bootstrapped, then certified | | **Who authors context** | Engineers, through the CLI, APIs, and the Workbench | Context Agents draft; domain experts approve inside [Bounded Context Spaces](https://atlan.com/context-engineering-studio/) | Decides who can fix a wrong definition | | **Context lifecycle** | [Context Cores](https://docs.trustgraph.ai/guides/context-cores/) package knowledge as offline, online, or loaded units; release process set by the operating team | Versioned Context Repos, simulated evals, approvals, rollbacks | Second-year cost sits here | | **Provenance and lineage** | [PROV-O extraction provenance](https://docs.trustgraph.ai/architecture/explainability) plus query-time reasoning traces | [Column-level lineage](https://atlan.com/data-lineage/) from SQL, pipelines, and BI; Iceberg time travel on past states | Two different audit questions | | **Retrieval** | Graph RAG, Ontology RAG, Document RAG, and [SPARQL](https://docs.trustgraph.ai/quickstart/sparql.html) | Knowledge graph traversal and vector-native search | Both go beyond flat vector RAG | | **Delivery to agents** | [REST, WebSocket, and Python APIs](https://docs.trustgraph.ai/reference/apis/); agents can register [external MCP tools](https://docs.trustgraph.ai/guides/mcp-integration/) | [Hosted MCP](https://docs.atlan.com/product/capabilities/atlan-ai/how-tos/atlan-mcp-overview), A2A, SQL, REST and Graph APIs | Both can reach an MCP-aware agent | | **Access control** | [Workspaces and a role model with reader, writer, and admin roles](https://docs.trustgraph.ai/architecture/security.html); external identity providers in enterprise tiers | Permissions enforced on every MCP call, set once in Atlan | Both scope what each agent can read | | **Operations** | Team runs the [storage and messaging components](https://docs.trustgraph.ai/architecture/storage.html) and inference | Managed by Atlan | Headcount versus subscription | --- ## How do TrustGraph and Atlan work together? Picture an insurer's platform team running TrustGraph over policy wordings, claims guidelines, and underwriting manuals. Its claims agent answers document questions well, with provenance back to the source page. When an adjuster asks for this quarter's commercial auto loss ratio, the agent has no certified definition, no path to the warehouse tables, and no way to know that finance and actuarial calculate the ratio differently. The scenario is illustrative, and the approach below would fix it. 1. **Keep TrustGraph on documents.** It continues extracting unstructured content into its hypergraph, with Ontology RAG and PROV-O provenance for document answers. 2. **Connect the estate to Atlan.** Atlan crawls the warehouse, dbt project, and BI tools, building column-level lineage from the claims tables to the loss ratio dashboard. 3. **Certify the definition.** Context Agents draft the loss ratio definition, Sage flags the finance versus actuarial conflict, and domain owners approve one version inside a Bounded Context Space. 4. **Test before shipping.** The certified definition ships in a versioned Context Repo, checked against questions simulated from the existing loss ratio dashboards. 5. **Give the agent both sources over MCP.** TrustGraph agents can [register external MCP tools](https://docs.trustgraph.ai/guides/mcp-integration/), so the claims agent can call Atlan's hosted MCP server for definitions and lineage. Teams using an orchestrator such as LangGraph can query both sources side by side instead. We found no documented direct integration between the two products, so the connection runs through MCP at the agent. TrustGraph's [MCP integration guide](https://docs.trustgraph.ai/guides/mcp-integration/) says production deployment and authentication for MCP may not be ready for production use. Atlan's MCP server connects with [OAuth or an API key](https://docs.atlan.com/product/capabilities/atlan-ai/how-tos/atlan-mcp-overview). Test the pairing on your own TrustGraph version before relying on it. With both in place, document questions resolve from TrustGraph with source-chunk provenance. Numeric questions resolve from Atlan with a certified definition, column-level lineage, and a recorded version of the context the agent used. One agent gets two audit trails that answer two different compliance questions. The [context layer implementation guide](https://atlan.com/know/how-to-implement-enterprise-context-layer-for-ai/) covers the Atlan side of this rollout step by step, and the [build vs. buy cost comparison](https://atlan.com/know/ai-agent/context-layer/context-layer-tco-build-vs-buy-vs-bundle/) helps when the choice is between extending a self-hosted graph and adopting a managed layer. --- ## Real stories from real customers building enterprise context layers ### How Workday is building an AI-ready semantic layer How Workday is building an AI-ready semantic layer "Atlan captures Workday's shared language to be leveraged by AI via its MCP server. As part of Atlan's AI labs, we're co-building the semantic layer that AI needs." Joe DosSantos, VP, Enterprise Data and Analytics Workday Workday's data, AI-ready Watch Now ### How DigiKey built a unified, sovereign context layer for its data and AI estate How DigiKey built a unified, sovereign context layer for its data and AI estate "Atlan is our context operating system to cover every type of context in every system including our operational systems. For the first time we have a single source of truth for context." Sridher Arumugham, Chief Data Analytics Officer DigiKey DigiKey's context operating system Watch Now --- ## Moving forward with Atlan and TrustGraph TrustGraph is a credible open-source foundation. For a document-heavy domain, a deployment your team wants to run end to end, or a product team embedding graph retrieval into its own application, it offers formal ontologies, three retrieval strategies, and provenance from source to answer. Setting up a context graph is the smaller part of the job. The larger part is keeping every definition, relationship, and policy true across a data estate that changes weekly, long after the engineers who wired it up have moved on. That is the work Atlan's enterprise context layer takes on: you own context for every domain team, with agents across the business reading the same certified definitions. --- ## FAQs about Atlan vs TrustGraph ### 1. What is the difference between Atlan vs TrustGraph? TrustGraph is an open-source context graph that engineers deploy to extract documents and data files into an ontology-grounded hypergraph, then retrieve from it with Graph RAG, Ontology RAG, or Document RAG. Atlan is an enterprise context layer that builds context from the live data estate through connectors, column-level lineage, SQL, and BI logic, routes it through human certification, and serves it to any agent. The core difference is where context comes from and who keeps it current. ### 2. Does Atlan replace TrustGraph? The two solve different problems, so many teams can run both. TrustGraph stays useful for document-centric retrieval and for deployments a platform team controls end to end. Atlan adds the warehouse and BI context, the certification workflow, and cross-agent delivery that a single-team graph leaves open. Teams that prefer one managed layer can consolidate on Atlan as the shared source of context. ### 3. What does Atlan add that TrustGraph does not list? As of October 2026, Atlan adds native connectors to warehouses, dbt, and BI tools, which do not appear on TrustGraph's integrations page. It also adds column-level lineage, nine Context Agents that bootstrap descriptions and ontology from estate signals, and versioned Context Repos with auto-generated evals and owner approval. Finally, it offers a managed Context Lakehouse with time travel for point-in-time audits and a hosted MCP server that needs no infrastructure to run. ### 4. Can Atlan read from TrustGraph? We found no documented direct integration between the two products, so the usual connection point is the agent, using MCP. TrustGraph agents can register external MCP tools, and Atlan runs a hosted MCP server that MCP-aware clients can reach with OAuth or an API key. TrustGraph's MCP guide says production deployment and authentication for MCP may not be ready for production use, so test the setup on your TrustGraph version first. ### 5. What is the biggest capability gap between TrustGraph and an enterprise context layer? Two gaps matter most. The first is source coverage: TrustGraph builds context from documents and files a team loads, while an enterprise context layer reverse-engineers it from warehouses, pipelines, and BI models where metric definitions live. The second is lifecycle: an enterprise context layer attaches ownership, approval, testing, and versioning to every definition, which a self-hosted graph leaves for the operating team to build. --- ## Sources 1. TrustGraph GitHub repository (Apache 2.0, README), TrustGraph, 2026. https://github.com/trustgraph-ai/trustgraph 2. TrustGraph homepage, TrustGraph, 2026. https://trustgraph.ai/ 3. TrustGraph 2.9 release notes, TrustGraph, 2026. https://trustgraph.ai/news/release-2-9 4. TrustGraph integrations, TrustGraph, 2026. https://trustgraph.ai/integrations/ 5. How does TrustGraph work?, TrustGraph documentation, 2026. https://docs.trustgraph.ai/learn/how-it-works 6. Ontologies, TrustGraph documentation, 2026. https://docs.trustgraph.ai/learn/ontologies.html 7. Explainability, TrustGraph documentation, 2026. https://docs.trustgraph.ai/architecture/explainability 8. Storage, TrustGraph documentation, 2026. https://docs.trustgraph.ai/architecture/storage.html 9. Security, TrustGraph documentation, 2026. https://docs.trustgraph.ai/architecture/security.html 10. Deployment, TrustGraph documentation, 2026. https://docs.trustgraph.ai/deployment/ 11. APIs, TrustGraph documentation, 2026. https://docs.trustgraph.ai/reference/apis/ 12. Context Cores, TrustGraph documentation, 2026. https://docs.trustgraph.ai/guides/context-cores/ 13. SPARQL quickstart, TrustGraph documentation, 2026. https://docs.trustgraph.ai/quickstart/sparql.html 14. Structured data load from a data file, TrustGraph documentation, 2026. https://docs.trustgraph.ai/guides/structured-processing/load-file.html 15. Custom processing, TrustGraph documentation, 2026. https://docs.trustgraph.ai/guides/building/custom-processing.html 16. MCP integration, TrustGraph documentation, 2026. https://docs.trustgraph.ai/guides/mcp-integration/ 17. GraphRAG: Unlocking LLM discovery on narrative private data, Microsoft Research, 2024. https://www.microsoft.com/en-us/research/blog/graphrag-unlocking-llm-discovery-on-narrative-private-data/ 18. Donating the Model Context Protocol and establishing the Agentic AI Foundation, Anthropic, 2025. https://www.anthropic.com/news/donating-the-model-context-protocol-and-establishing-of-the-agentic-ai-foundation 19. Atlan connectors, Atlan, 2026. https://atlan.com/connectors/ 20. Column-level data lineage, Atlan, 2026. https://atlan.com/data-lineage/ 21. Context Agents, Atlan, 2026. https://atlan.com/context-agents/ 22. Context Engineering Studio, Atlan, 2026. https://atlan.com/context-engineering-studio/ 23. Context Lakehouse, Atlan, 2026. https://atlan.com/context-lakehouse/ 24. Atlan MCP overview, Atlan documentation, 2026. https://docs.atlan.com/product/capabilities/atlan-ai/how-tos/atlan-mcp-overview 25. Atlan Application SDK (Apache 2.0), Atlan, GitHub, 2026. https://github.com/atlanhq/application-sdk 26. What Is a Context Graph? The AI Reasoning Layer Explained, Atlan on YouTube, 2026. https://www.youtube.com/watch?v=cTfuzlcrIMQ