What's shipped. What's in flight. What's next.
Eisberg is the agentic OS that retires the warehouse era — the substrate every function in your company (not just the 10% who write SQL) configures and runs on. The eight primitives below are what every team uses to do that. Connectors, agents, metrics, workflows — we ship some as defaults, but the goal is that customers build their own through no-code Builder surfaces. This page is honest about what's there today vs what's planned. Honesty about gaps is the only way the dethronement claim holds up under scrutiny.
Connectors
In flightConnect to any source — operational systems, accounting platforms, CRMs, payment + marketing automation, vertical SaaS. ~30 connectors today; SDK lets anyone build more.
Customer customization
Connector SDK (Python ABC, 7-hook lifecycle: introspect → catalog → ontology → CDC → classify → policy → lineage)
Storage
ShippedOpen Apache Iceberg in your own cloud bucket, behind your KMS keys. Customer-owned, structurally portable, any compliant engine can read it.
Customer customization
Customer-managed — Eisberg never holds records
Engines
ShippedMulti-engine query router. Embedded for sub-second micro-queries, distributed for medium, GPU-native for heavy aggregation, MPP-on-Iceberg for real-time at scale. Routed per workload.
Customer customization
Engine flag per workspace; new engines slot in behind one interface
Semantic layer
ShippedOSI v1.0 native import + export. Metrics defined in dbt, Cube, ThoughtSpot, Sigma, or any OSI-conformant tool travel into Eisberg without translation.
Customer customization
Metric Builder UI (no-code) — planned Q3 2026
Agents
ShippedEvery autonomous agent receives a signed Birth Certificate at creation — identity, intent scope, data class allow-list, action ceiling, expiry, human owner, policy bundle hash. Job spine refuses to compile out-of-scope actions.
Customer customization
Agent Builder UI (no-code) — planned Q3 2026; Birth Certificate SDK as embeddable primitive — planned Q3 2026
Governance
ShippedPolicy-as-code at every layer (RBAC + ABAC + RLS + dynamic masking + temporal + consent + purpose limitation). Column-level lineage with cryptographic chaining. Audit packs in under 60 seconds.
Customer customization
Custom policy definitions per workspace; pre-built compliance modules for SR 11-7 / BCBS 239 / HIPAA / 21 CFR Part 11
Workflows
ShippedUnified Job spine — any intent compiles to a workflow with budget, observers, self-healing, approval gates, durable execution. Survives infrastructure failures.
Customer customization
Workflow Builder UI (no-code) — planned Q4 2026
Ontology
ShippedAuto-linker proposes entity relationships; humans approve. Built incrementally from your data, not from a consulting whiteboard. Six months of consulting replaced by two weeks of watching.
Customer customization
Ontology editor — proposes via auto-linker, accepts manual edits
22 shipped. 6 in flight. 10+ planned. SDK lets anyone build more.
The most-asked-about surface. Honest tally below. The strategic bet: we ship the connectors that gate the highest-leverage named accounts, and the Connector SDK lets customers + partners ship the long tail. The single biggest move on the roadmap is the generic Airbyte adapter — one build unlocks Airbyte's 700+ sources without writing 700 connectors.
22
- Snowflake
- Databricks
- PostgreSQL
- MySQL
- MongoDB
- Oracle
- SQL Server
- Teradata
- Streaming CDC pipeline
- Salesforce
- HubSpot
- Stripe
- Shopify
- QuickBooks Online
- Klaviyo
- Procore
- Toast POS
- S3 / cloud storage
- SFTP
- REST API (generic)
- Google Sheets
6
- Bullhorn (staffing ATS)
- Olo (restaurant)
- 7shifts (restaurant labor)
- Sage Intacct
- NetSuite
- Recharge (subscription)
10+
- Applied Epic + Vertafore (insurance)
- Acumatica
- Plex (manufacturing)
- Workday
- Microsoft Dynamics 365
- Sage 300 CRE
- Viewpoint Vista
- ADP / Paychex / Gusto (payroll)
- EDI adapter (SPS Commerce / Cleo)
- Generic Airbyte adapter (unlocks 700+ Airbyte sources via one build)
Don't see your system? The Connector SDK lets anyone ship a connector in 2-5 days. Available to design partners + open-sourced at GA. Active design-partner pilots get any connector they need accelerated as part of the pilot.
The wedge nobody else has — legacy code, stored procedures, macros, email, comms, docs — ingested as governed facts.
The actual rules your company runs on are not in your data warehouse. They are buried in a 1998 stored procedure that excludes vendor 4711, an Excel VBA macro that computes DSO with a 90-day cutoff, an ABAP routine that classifies revenue recognition, an email from the controller in 2019, a Slack thread, a Confluence page nobody updates. Eisberg's tribal-knowledge connectors parse the code body, the macro body, the message body — extract entities via LLM, bind them to your auto-discovered business ontology, and serve them back as first-class governed facts cited in every agent answer. Status of each source below.
0
- None yet — first wave in flight below.
5
- Slack workspace + channel ingestion
- Confluence space + page ingestion
- GitHub / GitLab PR bodies + code comments + ADRs
- Stored procedures + SQL views + triggers (SQL Server / Oracle / Postgres / Snowflake)
- Excel formulas + VBA macros + named-range extraction
10+
- Legacy code parsers — COBOL, MUMPS, RPG, SAS, Fortran, ABAP, PowerBuilder
- SSIS packages + Informatica mappings + dbt models + SAS programs
- Email archives (Gmail vault / Outlook PST / Microsoft 365 mailbox)
- Microsoft Teams + SharePoint
- Notion + Google Docs + Office 365 + OneDrive
- Meeting transcripts (Recall.ai / Otter / Gong / Chorus / Read.ai)
- Internal wikis + runbook stores
- Audit walkthroughs + SOX evidence narratives + regulatory filings
- Support ticket histories (Zendesk / Intercom / Salesforce Service Cloud)
- Sales call libraries (Gong / Chorus)
Tribal knowledge ingestion passes through the same seven-policy plane and the same Birth Certificate scoping that protects structured data. Consent is per-source, revocable per-source, time-bounded. Your tribal knowledge lives in your bucket — Eisberg never stores it on Eisberg infrastructure.
What customers configure themselves — no engineering ticket required.
The bones of Eisberg (storage, engines, governance, semantic layer, agents, workflows) are general-purpose primitives. The work ahead is making each primitive customer-configurable through no-code Builder surfaces. This is where the agentic OS thesis lives: every team builds the agents + metrics + workflows + dashboards their work actually needs.
Connector SDK
ShippedPython lifecycle ABC + 7-hook chain. Anyone with API + DB knowledge can ship a custom connector in 2-5 days. Open spec; we publish reference connectors as examples.
Birth Certificate SDK
In flightStandalone Python + TypeScript package for AI-native SaaS to embed agent governance in THEIR product. Same primitive we use internally, packaged for external use. Closes EU AI Act Article 11 (effective Aug 2026) for their procurement.
OSI v1.0 metric import/export
ShippedBring metrics from dbt Fusion, Cube, ThoughtSpot, Sigma, MetricFlow, or any OSI-conformant tool. Re-export back to any tool. Zero translation.
Agent Builder UI (no-code)
PlannedDefine your own agent through the workspace UI — intent type, scope, allowed data classes, action ceiling, expiry, human owner. Signs Birth Certificate. Compiles into the Job spine. Planned Q3 2026.
Metric Builder UI (no-code)
PlannedDrag-drop dimensional model, exports as OSI v1.0. Validates against your ontology + lineage. Planned Q3 2026.
Workflow Builder UI (no-code)
PlannedVisual graph editor — drag steps, wire data, set approval gates. Compiles to the unified Job spine. Planned Q4 2026.
Dashboard Builder via natural language
Planned"Show me churn risk by segment for the last 12 months" → live dashboard, governed by your policy bundle, citing every source. Planned Q4 2026.
Document Builder UI (no-code)
PlannedFor regulated industries — insurance underwriting, pharma regulatory + PV, medical devices (510(k) / PMA / CTD), legal contracts, public-sector benefits, GMP/GLP/GCP quality. Template-driven recognition + section extraction + cross-reference resolution + structured authoring. Adds the regulated-document workflow primitive. Planned Q3 2026.
Schedule Builder UI (no-code)
PlannedFor calendar-shaped work the DAG editor can't express — payroll cycles, financial close, OR scheduling, clinical visit cadence, dispatch routes, billing cycles, audit cycles, renewal cycles. Gantt + dependency + resource-pool view. Planned Q4 2026.
Action Builder UI (no-code)
PlannedClosed-loop write-back to operational systems with policy gating + audit trail. The last mile of agent decides → action lands in Salesforce/SAP/Workday with a signed audit record. Planned Q4 2026.
Knowledge Layer surface (`/knowledge`)
In flightUser-facing dashboard showing what the platform has learned — auto-discovered ontology, query patterns, classifications, tribal facts ingested per source, persona-specific recent activity, federated patterns received under k≥3 anonymity. Read-only v0.5 in flight; full ontology editor follows.
Examples customers start from — not the product.
We ship reference configurations for HR, Marketing, Engineering, Supply Chain, Finance, and Customer Success today. Five more are planned (Compliance, Construction, Manufacturing, Restaurant operations, RevOps / Lead-to-Cash). These are NOT 'vertical modules' you choose from like Salesforce clouds. They're working examples of agents + metrics + workflows you fork and modify. Your team builds whatever you actually need.
What a starter pack contains
- 5-10 named entities with ontology relationships (e.g., for Finance: Account, Invoice, Vendor, Payment, Journal Entry)
- 10-20 named metrics in OSI v1.0 format (e.g., Days Sales Outstanding, Operating Margin, Cohort Retention)
- 3-5 starter agents with Birth Certificates already scoped (e.g., Variance Investigator, Cash Flow Risk Detector)
- 2-3 compliance modules mapped (e.g., SOC 2 evidence collection, SOX-adjacent journal review)
- Reference dashboards a human can read; reference workflows a human can extend
Every pack is a starting point. The platform doesn't lock you into the pack — your team forks, modifies, or replaces every piece. The packs are open by design; nothing inside them is special-cased in the platform.
The honest exclusion list.
Per our `gtm/research/08-data-lifecycle-competitive.md` analysis, four categories are knife-fights we lose if we try to build them. We partner instead.
We don't build
BI / Visualization
We feed these tools your unified Iceberg substrate. Partner stack: Looker, Sigma, Mode, Hex, Tableau, Power BI, Metabase, ThoughtSpot.
We don't build
Customer Data Platform (CDP)
We feed these tools your unified Iceberg substrate. Partner stack: Segment, RudderStack, mParticle.
We don't build
Product Analytics
We feed these tools your unified Iceberg substrate. Partner stack: Mixpanel, Amplitude, PostHog, Heap.
We don't build
Marketing Attribution
We feed these tools your unified Iceberg substrate. Partner stack: Triple Whale, Northbeam, Rockerbox.
Want any of these accelerated?
Design-partner pilots get any connector or builder surface in the planned column prioritized as part of the pilot scope. If your sign-off depends on a specific system or workflow, we build it in the pilot — at no additional cost. That's the deal.
Where to go next.
Platform overview
The eight primitives, the six moats, the architecture properties.
See the platform
vs Snowflake
Snowflake gives data engineers tables + SQL. Eisberg gives every team primitives + builders.
Read the comparison
Agent Trust
Birth Certificates, durable execution, governance defined at creation — the trust primitive every agent inherits.
Read trust deep-dive