Agenticorg

AgenticOrg

AgenticOrg is an enterprise AI agent platform for building, running, and governing agent workflows across finance, HR, marketing, operations, back office, communications, and bounded agentic-commerce use cases. The platform is source-available under the Orchestrum Personal Use License 1.0: only individuals' personal use is free, and any organisation needs a signed agreement with Orchestrum Technologies LLP; the SDKs and MCP server are Apache-2.0 open source (see Licence).

Ownership: AgenticOrg is owned by Orchestrum Technologies LLP.

Inventor / Owner: Sanjeev Kumar

Contact: sanjeev@orchestrum.in | mishra.sanjeev@gmail.com

Website demo and trial lead notifications default to sanjeev@orchestrum.in. Self-hosted operators can select their own single marketing inbox with AGENTICORG_DEMO_NOTIFY_TO. This does not redirect requester confirmations, workspace notifications, or account-recovery email, and does not replace SMTP setup.

See the canonical ownership and contact statement and NOTICE.

Website | Application | Playground | User manual | Technical documentation | Current product status | Security policy

Start with a measurable pilot: bank service cases, fintech/API integrations, or merchant and external-buyer workflows. The visual rollout checklist explains source setup, review, external confirmation and expansion. Discuss your task and systems at sanjeev@orchestrum.in; do not send credentials or customer data in a public enquiry.

Licence

The AgenticOrg platform is source-available under the [Orchestrum Personal Use License 1.0](LICENSE). Only an individual using it for their own personal purposes is free. Every organisation (companies, firms, sole proprietors' businesses, charities, schools, universities, research bodies and government) needs prior written permission in a signed agreement with Orchestrum Technologies LLP, for any use: evaluation, testing, pilots, internal use, hosting, resale or embedding, paid or free.

The AgenticOrg SDKs (`sdk/`, `sdk-ts/`) and the MCP server (`mcp-server/`) are open source under the Apache License 2.0, free for any use including commercial.

• Free (platform): an individual's own personal use only. Distributing the platform needs a written agreement too.
• Commercial licence: see COMMERCIAL-LICENSE.md and write to sanjeev@orchestrum.in.
• Existing users: these terms apply to every version published on or after 21 October 2026, including all updates. Versions published before that date remain available under the Apache License 2.0 they were released with. Existing and new users alike need a signed agreement for any use by or for an organisation of a version from 21 October 2026 onward.
• Trademarks: "AgenticOrg" and "Grantex" are trademarks of Orchestrum Technologies LLP; commercial use of the names, logos, hosted services or brand needs prior written approval with any version. See TRADEMARKS.md.

Source of truth

Live product totals and the deployed product version are published by the public product facts endpoint:

GET https://app.agenticorg.ai/api/v1/product-facts

The response contains version, agent_count, connector_count, and tool_count. Do not copy those values into documentation, page copy, or tests: they are computed from runtime registries and can change independently of a documentation release. The version declared for source builds is in pyproject.toml.

Billing limits and list prices come from core/billing/catalog.py. Production deployment behavior comes from scripts/deploy_cloud_run.sh.

Current release at a glance

Bring your own channels and AI credentials

New agents default to gemini-3.5-flash-lite; explicitly selected models remain unchanged. Save your Google or other provider key in the protected AI credential settings. Model selection, fallback and company scope remain configurable. For this model, Agent Configuration > Gemini thinking offers an optional medium profile for multi-step reasoning. Existing agents and generic calls keep the provider default; no global flag or default is changed. The API equivalent is llm.thinking_level="medium"; set it to null to restore the provider default. Both runtime clients honor explicit profiles. Thinking is billed as output tokens; budget for its additional token use and response time within existing limits. It is not a correctness guarantee. Older model settings and all feature flags remain unchanged. See reasoning acceptance.

The Channels dashboard connects your own WhatsApp Cloud API, Meta Messenger, Telegram, Viber, inbound Mailgun email or signed gateway to an agent or workflow. Credentials are encrypted, callbacks are authenticated, unlinked messages require review, and private sender links delegate only your current permissions. Existing Twilio voice/call setup and remote MCP tool discovery remain separate integrations. Native channel replies default to optional status notices. Telegram agent connections can opt into bounded final-answer replies for the linked private sender under Channels > Reply content. Traces and tool-call records stay in AgenticOrg; approvals, sender authorization and tenant controls still apply. The encrypted intake carries the authenticated sender identity across services; older receipts without that proof remain withheld, even if the agent completed. For tool-free Telegram agents without traffic splits in this opt-in mode, follow-ups may use up to four recent, confirmed question/answer pairs from the same linked private sender and agent/company scope, within 30 minutes. This is bounded text context, not durable memory or permission to repeat actions. Hindi and other Unicode model input remains readable before privacy scanning; existing redaction still applies. With AGENTICORG_PII_HINDI_GRAMMAR_ENABLED=true (default off), targeted Hindi grammar checks distinguish referential name/colour questions from named values. This is not a general-purpose Hindi entity-recognition model; quoted names, unknown words and structured identifiers retain privacy scanning.

Follow the channel account and workflow guide for account creation, callback registration, sender linking and verification. Local wire-contract tests are not proof of real provider-account delivery. Reuse an existing Telegram connection for an already connected bot; a second connection must not overwrite its webhook. Change its destination through Save settings only when authorized and no tasks are queued or running. Verify real model reasoning separately from transport-only workflows, and do not claim a combined Telegram-to-Gemini flow unless one delivery proves both.

Acceptance checks validate the actual answer as well as the terminal status. A completed, high-confidence model run can still be wrong; compare numeric results independently and preserve PII protections while diagnosing it. See the bounded production acceptance runbook.

The production platform currently includes governed agent/workflow execution, document ingestion with scanned-document OCR, signed Twilio voice with provider-managed STT/TTS, tenant-scoped browser RPA, Python and TypeScript SDKs, MCP/A2A surfaces, and the bounded Shopify-to-OACP commerce runtime. See the current product status for the exact shipped, configuration-dependent, and not-shipped boundaries.

External actions remain explicit. A voice provider, connector, browser script, buyer channel, payment rail, merchant system, or POS path is usable only when the target tenant has the required credentials, scopes, configuration, and provider approval.

Governed business cases

The repository includes a tenant-gated business onboarding case workflow with read-only provider investigation, a cited memo, deterministic policy scoring, screening dispositions, human review, and an auditable hand-off. Provider calls are refused unless the active role agent, registered tools, exact local case purpose, and delegated grant all authorize the call. This case-specific check is strict even if general tool enforcement is in off or warn mode.

AgenticOrg pins grantex==0.7.1 and calls its enforcement check with the grant token, connector, tool and any numeric amount. It does not pass a case id or case purpose, so a token's case purpose and per-case caps are not checked on this path. The role's case_purposes list is a local AgenticOrg control, and pooled grants are not bound to one case. Decision and analyst actions still require a signed-in human; API keys, agent tokens, and MCP tools cannot act as that human. The case runtime is off for a tenant until enabled and configured. This source description is not a claim that a particular hosted deployment has enabled the workflow. See the case lifecycle and setup, grant-enforcement runbook, and decision requests.

Voice, email, OCR, and browser RPA can be exercised together before release with the Docker-backed local multichannel simulation. It uses fresh PostgreSQL, local Mailpit delivery, real Tesseract, and optional real Playwright Chromium while keeping paid calls and production mutations off.

Knowledge upload status distinguishes extraction from searchable indexing. With opt-in upload recovery, transient native embedding failures retry on the existing worker; the previous searchable copy remains until replacement succeeds. Malformed PDF/Office files are refused before replacement. See Knowledge and OCR for status meanings, citations and recovery limits. For a reasoning-only agent, disable suggested tools and leave tool/connector selection empty; this does not bypass tenant budgets. Provider charges and quotas still apply. Use structured output for conf