What it does
7 features
Langdock · AI platform for companies with agents and workflows
Product and access
Langdock is a proprietary product of Langdock GmbH in Berlin. The platform combines models from several providers; agents and workflows are parts alongside chat, meetings and apps.
Core features
7 features from provider sources · checked on 5 November 2026. Availability and limits are directly below.
Agents with knowledge and actions
An agent combines instructions, a model, up to 50 files and actions through integrations, such as updating CRM entries or creating tickets. The documentation describes agents as specialised chatbots with tools.
Source 12Workflows with triggers and nodes
Workflows connect agents, integrations and custom logic into multi-step processes with conditions and loops. They start from a form, a schedule or an event in a connected app.
Source 13Scheduled tasks
A saved prompt runs daily, on weekdays, weekly or monthly, optionally with an agent. Each user can have 10 tasks per workspace, and starts can be delayed by up to 15 minutes.
Source 21Integrations and MCP
The page lists 57 integrations and 754 native actions. External MCP servers can be connected; up to 60 tools and 50 resources can be enabled per integration.
Source 17Approval steps in workflows
The "User input" node pauses a run until someone with editor access approves, rejects or submits a form. The run then continues on the matching branch.
Source 15Governance for agents
According to the page, admins can approve, flag or disable agent versions. Sharing with a team or the whole workspace can require approval.
Source 7
Show all 7 features
Deployment options
The default is multi-tenant SaaS on Microsoft Azure in the EU. According to the page, single-tenant starts at 2,000 seats, and your own cloud or on-premise via Helm charts at 5,000 seats.
Source 3
Usable AI models
Open choice
By default you choose from Langdock's list of several providers. With your own keys (BYOK) you can also connect any OpenAI-compatible endpoint, such as LiteLLM, vLLM or Ollama; you select BYOK when moving to a paid plan or later through the Langdock team. The endpoint must be publicly reachable over HTTPS; internal addresses only work through support.
Model families named
- Google (Gemini)
- OpenAI (GPT)
- Anthropic (Claude)
- Meta (Llama, Muse)
- Mistral
- DeepSeek
- Other models can be connected freely
- Model choice
- You choose
- Your own API key
- Yes
- Router or custom endpoint
- Yes
- Local models
- Yes
- Depends on
- Plan, Admin approval
The rating covers the generally available main access. Documented · Sources accessed on
Langdock: Models · Langdock: BYOK · Langdock: BYOK setup guides
Does it fit my task?
2 practical tasks · 6 requirements and limits
A good fit if …
- you want chat, agents and workflows with several models on one platform for the whole company
- processes should run on a schedule or event and wait for human approval at fixed points
- admins should control agent versions and connections centrally
Less suitable if …
- you want to check the sub-processors on a freely readable page before signing
- your company is well below 5,000 seats and wants to operate the platform itself
- agents should remember things across sessions through memory
Editorial assessment based on provider documentation. Not a product test.
Practical tasks
Editorial examples based on documented features. No product tests performed.
Pre-sort incoming email
- Your input
- A workflow with an event trigger, an agent node and an approval step.
- Possible result
- The agent classifies the email and creates a draft reply that waits for approval.
- What to check
- Review inputs and outputs per node in the run and choose the error option for each node deliberately.
Weekly briefing
- Your input
- A scheduled task with a selected agent and its knowledge.
- Possible result
- Each run creates a new chat with the summary, which you can open and continue.
- What to check
- Check whether actions need confirmation, because a waiting run counts as "Failed".
Use a workflow to classify incoming support emails and present a draft reply for approval.
Your possible prompt
Compare three providers for [task]. Use their official documentation. Show differences, sources, and open questions in a table.
Adjust the task, data and permissions.
Requirements and limits
Requirement
Workflows need an admin and a plan
An admin must enable workflows in the workspace settings. According to the pricing page they require an active Core or Work plan for all users, and they use AI credits.
Source 2Limitation
Memory does not apply in agent chats
Memory is available only in regular chats, not in project or agent chats. It holds at most 50 entries.
Source 16Limitation
MCP tools without default confirmation
Newly saved MCP tools do not require confirmation by default. You have to enable it for each tool afterwards.
Source 17Limitation
No built-in MFA
According to the documentation, Langdock has no standalone MFA. With SAML, MFA is enforced in the identity provider.
Source 19Not yet confirmed
Sub-processors not readable
The privacy policy points to the Trust Center for the list. The Trust Center returned only its title when fetched, so we could not check the names.
Source 10Not yet confirmed
Model region labels
The models page marks models with an EU or globe symbol, for example Gemini 3.5 Flash and DeepSeek v3.1 with a globe. It does not explain what the globe symbol means for data processing.
Source 8
Can I use it?
With requirements
Europe & regional availability
Security page, models page, privacy policy and documentation reviewed. The Trust Center with the DPA, certificates and sub-processors returned only its title when fetched.
Reviewed scope: Hosted platform in the EU; models depending on selection
Reviewed on:
Available
Hosting in the EU
According to the security page the multi-tenant SaaS runs on Microsoft Azure on servers in the EU. According to the page the application and "most" models are fully hosted in the EU, configurable.
Region, reason and source
Region covered: EU
Documented reason: Provider statement; the page does not specify the Azure data centre further.
With requirements
Models outside the EU label
The models page shows models from OpenAI, Anthropic, Google, Meta, Mistral and DeepSeek with an EU or globe symbol. GPT-5 Pro, Llama 4 Maverick and DeepSeek v3.1, for example, carry a globe. With BYOK you use your own keys with providers such as Azure, Anthropic or Google.
Region, reason and source
Region covered: EU and beyond
Documented reason: Processing depends on the model chosen; the page does not explain the globe symbol.
Official source · www.langdock.com · Official source · docs.langdock.com
Available
Data processing agreement
According to the documentation the data processing agreement is part of the terms and applies automatically with use. The document itself is in the Trust Center and could not be read.
Region, reason and source
Region covered: EU/EEA
Documented reason: Existence of the agreement is documented; its content was not reviewed.
Official source · docs.langdock.com · Official source · trust.langdock.com
Availability still unclear
Sub-processors and third countries
The privacy policy allows transfers to third countries outside the EU and EEA with suitable safeguards and points to the Trust Center for sub-processors. We could not read any names.
Region, reason and source
Region covered: EU/EEA
Documented reason: Trust Center not readable when fetched.
Official source · www.langdock.com · Official source · trust.langdock.com
Availability still unclear
Certificates and "GDPR compliant"
The security page mentions ISO 27001 and SOC 2 Type II with independent audit as well as "GDPR compliant". According to the documentation the certificates are in the Trust Center. "GDPR compliant" is a provider claim with no audit report on the page.
Region, reason and source
Region covered: EU/EEA
Documented reason: Certificates not seen; no authority assessment found.
Official source · www.langdock.com · Official source · docs.langdock.com
Accounts & permissions
Based on provider documentation. Permissions depend on the account, integration and organization policies. Review the permissions actually requested before use.
Configuration covered: Hosted platform with workspace administration
Sources reviewed:
Show all 3 points
Documented with conditions
Data access & team sharing
OAuth connections act with the rights of the signed-in person by default. Through an agent with a preselected connection, others can run the action with those credentials. Application permissions can reach across the organisation; the documentation advises a clear need.
Evidence
Memory and conversation continuity
3 of 6 criteria documented
Memory stores up to 50 facts and preferences in the account but does not apply in agent chats. Agents work with files and folders.
Chat memory for regular chats; agents use instructions and knowledge
Sources accessed:
Documentation by criterion
StoragePartly documented
Storage and scope
According to the documentation memories are stored in the account, at most 50 at a time. The page does not name the physical storage system.
SelectionDocumented
Selection and timing
The model recognises what is worth keeping, and you can explicitly ask it to save something. Very personal, short-lived and sensitive details are saved only on explicit request.
Edit & deleteDocumented
Inspect, edit and delete
In the Memory tab of the account settings you view, edit and delete entries. You can also do this by instruction in chat.
Access boundariesDocumented
Agent and channel boundaries
Memory applies only in regular chats, not in project or agent chats. Workspace and personal settings must allow it, and the model must support tools.
RetrievalPartly documented
Retrieval
Saved entries are available in later regular chats. The documentation does not say how retrieval works technically. Agent knowledge is retrieved as relevant excerpts.
New conversationsPartly documented
Session reset and compaction
According to the documentation chat history can be summarised over time. It does not describe resetting memory beyond deleting entries.
Cost and effort
4 plans · Prices as of 5 November 2026
Prices and included usage
Pricing checked
Prices per seat per month excluding VAT; the tiers on the page start with the first 50 seats. The trial lasts 7 days, without a credit card and with EUR 5 of model credit.
| Plan | Price | Included |
|---|---|---|
| Core | EUR 23.20 per standard seat, EUR 99.00 per Core Max seat | Chat, agents, skills, integrations, SSO, SCIM and SAML, 2,500 workflow runs per month, 10 transcribed meetings per member. Governance until 1 January 2027 according to the page. |
| Work | EUR 31.20 per standard seat, EUR 109.00 per Work Max seat | Everything in Core, plus governance, unlimited meetings with a fair-use limit of 60 hours per week and 10,000 workflow runs per month. |
| Enterprise | Custom, from 1,000 users | Volume pricing, dedicated deployment possible, individual support. |
| Workflows Business | Add-on: the page lists EUR 539 per workspace per month as well as EUR 449 for 40,000 and EUR 999 for 100,000 runs | More workflow runs. Unused runs expire at the end of the billing period. |
- The page lists several amounts for Business side by side and does not explain how they relate.
- The home page mentions custom deployments from 1,000 seats, the security page from 2,000 and 5,000 seats.
- The API is billed separately by tokens. Third-party price information was not used.
Setup
Start a trial without a credit card. For agents with actions you set up connections, and an admin must enable workflows.
Operation & oversight
Based on public provider documentation. Assessments are editorial inferences from the stated properties. Actual time requirements and reliability have not been measured.
Deployment covered: Hosted operation by the provider; self-operation only from large seat numbers
Sources reviewed:
Editorial assessment
Langdock documents triggers, approvals and error options for workflows in concrete terms. The sub-processor list and general retention periods remain open. For MCP tools you should enable confirmations yourself.
Show all 6 points
Documented with conditions
Files and working state
Agents use up to 50 files, 5 library folders and 5 synced folders. According to the pricing page workflows keep data for 30 days in the Starter scope. The reviewed pages name no other periods; according to the security page data stays as long as needed for the service.
Evidence · Evidence · EvidenceDocumented with conditions
Resuming after failures
Per node you choose stop, continue with the error as output, an error path or retry. Up to 5 retries are possible, but not for User input, Delay, Loop and triggers. The reviewed page does not describe resuming aborted runs.
Evidence
Evidence and open questions
23 sources · 4 open points
Open points
The reviewed sources contain no reliable information on these points. A gap says nothing about the product.
Europe: Availability still unclear
Requirements and limits: not yet confirmed
Sources and evidence
Official website, pricing page, security page, privacy policy and documentation reviewed. The Trust Center showed only its title when fetched. Feature descriptions come from the provider; example tasks are editorial. Our own hands-on tests are still pending.
Editorial team: Michael Meierhoff · sources reviewed on . Provider information is not independent evidence of performance.
Provider
Langdock GmbH, based in Berlin, offers an AI platform for companies.
- Operator
- Langdock GmbH
- Organization
- Provider
- Website
- www.langdock.com
Changes to this profile 1
Commercial AI platform from Berlin with chat, agents, workflows and integrations for company-wide use.
These notes describe editorial changes. Source and pricing review dates appear in their respective sections.