With review
The agent prepares materials; you approve them in Kontentika before publishing. Other actions depend on the granted permissions.
CONNECT AN AI AGENT
Your AI agent.
Your social channels.
One place to manage them.
Connect your AI agent once, then manage social work by talking to it. It can maintain brand knowledge, prepare content, schedule posts, reply to your audience and read analytics. Autopilot removes per-post review; you choose the accounts, permissions and limits.
- No social-network passwords shared
- Revoke access any time
How work flows
Your AI agentGive instructions in your usual chat
KontentikaOne control centre
Your permissions · Your limits
- Content and brand knowledge
- Publishing and calendar
- Messages and analytics
- Paid advertising
Actions and results stay in Kontentika
Choose how independently your agent works
Two modes. Your choice.
Autopilot
The agent prepares, edits, schedules and publishes, replies to your audience and reads results within its permissions. You do not need to approve every post.
Connect your accounts
Sign into each social platform and allow Kontentika to work with it. Enter passwords and security codes only on that platform.
Grant working access
Open AI agents, select Autopilot, the workspace and accounts. Choose individual permissions or Select all work permissions.
Set boundaries
Set daily material, publication, reply and Kontentika-generation limits. 0 disables our generation. Add standing instructions and choose a duration or Until revoked.
Confirm autonomous publishing
Check the separate publishing-without-review confirmation and save. Existing connections never gain extra rights automatically.
Give your agent a task
For example: Run Telegram and Instagram, publish three useful posts weekly, answer routine questions, escalate disputes and send me a weekly report. The agent must check actual permissions and platform support.
You may still need to complete initial social sign-in, renewed provider consent and security checks. The agent receives no social passwords, cannot change payment details or administer users. Advertising requires separate permissions and a financial grant.
Accepted work continues with chat closed
Kontentika persists accepted preparation, media and publication jobs. You can submit a complete background campaign. For an external agent to invent new tasks or react independently to incoming messages, its client needs background or recurring execution. An MCP connection alone does not make every chat a 24/7 worker.
From content to promotion
Your agent can manage paid advertising too
Ask your agent to prepare campaigns, check results and manage delivery. Choose advertising cabinets, currency and budgets separately from social-content access.
Prepare and approve
The agent uses your materials to prepare creative, targeting, dates and budget. In Review mode, an owner or admin approves each campaign.
Launch within the grant
Advertising Autopilot can approve and create a paused campaign, then launch after moderation checks. A shared budget covers all campaigns allocated to this connection.
Monitor and adjust
The agent reads spend and results, pauses campaigns and adjusts budgets within your grant. Increases require enough data and a waiting interval; uncertain operations are never retried blindly.
Pausing does not release allocated budget. Network reporting is delayed: settings and reported spend are checked, but exact billing cannot be guaranteed. Features depend on the connected network; unsupported external campaign hierarchies must be launched in the network’s Ads Manager.
You can separately stop the agent’s advertising and block new spending. Unconfirmed provider stops remain visible.
Set up advertising access ↗Start with a connection
Choose how to connect
You need an active Kontentika plan and workspace owner/admin access. For Autopilot, first connect your social accounts, then authorize the agent for the accounts you choose.
NO KEY TO SHARE
Sign in through Kontentika
For agents that open a sign-in window and ask you to approve access. We tested this flow in ChatGPT.
Sign-in steps ↓WITH AN ACCESS KEY
Create an access key
For agents with a secret key field in their connection settings. Create a separate key in Kontentika first.
Access key steps ↓Actual application screens with demo data and the live server address. Keys are masked. Select an image for details.
NO KEY TO SHARE
Sign in through Kontentika
https://app.kontentika.app/mcp
Open a step to view the screen
Copy the connection address
Open “AI agents” in your account. Under “Connect an agent”, select “Copy” next to the server address.
Copy the connection address Add Kontentika to your agent
In ChatGPT open “Plugins” → “Add” → “Add custom MCP server”. Name it Kontentika, paste the address above and choose OAuth authentication. Review permissions, create the personal plugin, then select “Connect Kontentika”.
Add Kontentika to your agent Choose a workspace and permissions
Choose the operating mode and permissions. Reading is enough for an initial check. Autonomous publishing requires Autopilot and a separate confirmation.
Choose a workspace and permissions Set limits and approve
Select accounts, duration or access until revoked, daily limits and standing instructions. Our generation is disabled by default. Select Allow access.
Set limits and approve Check the connection
Ask the agent to read the workspace name and drafts. A row appears in “Your connections” and “Last used” updates. Test reading before allowing further actions.
Check the connection
WITH AN ACCESS KEY
Create an access key
Open “AI agents”
Select “Create access key” under “Your connections”. Give each agent a separate connection.
Name the connection and choose permissions
Name the connection and choose workspace, mode and rights. For Autopilot select connected accounts and separately confirm publishing without review.
Name the connection and choose permissions Set the duration and caps
Set limits and instructions. Give our generation a positive daily cap if needed; 0 keeps ready-content imports available. Choose an expiry or access until revoked.
Set the duration and caps Save the key in your agent settings
On “Copy this key now”, select “Copy”. Paste it into the agent’s secret connection field, then select “I saved the key”. It cannot be shown again. Never paste it into an ordinary chat or address bar.
Save the key in your agent settings Check reading and last use
Give the agent the server address and ask it to read the workspace. Check “Last used” in “Your connections”. If there are no calls, review the client settings instead of guessing at broader permissions.
TESTED IN A REAL CLIENT
ChatGPT: connection and reading verified
On 24 October 2026 we tested a personal ChatGPT plugin: Kontentika sign-in, permission approval, workspace_context and list_materials calls. We granted read-only access to a test workspace for one day and revoked it afterwards.
Tested on 24 October 2026 in ChatGPT on the web, on the Pro plan, with the “Plugins” menu. Menu names can change: check the official OpenAI documentation below. Connection availability depends on your account and workspace policies. Claude and Cursor are not yet listed as tested. The ChatGPT check covered connection and reading. Background execution and action confirmations depend on the client.
Official OpenAI documentation for connecting ChatGPT ↗First connection check
Use Kontentika to read my workspace name and show one draft. Do not create, change or publish anything.
When something goes wrong
The agent receives a failure cause, stage and safe next action. Recovery is enabled separately in Agent connections.
Check the campaign first
Before launch, Kontentika checks the workspace owner subscription, daily and monthly allowances, generation configuration and selected connections. This check creates no content and spends no credits. It does not guarantee remote platform availability.
Keep completed work
The agent can resume a failed material or required media from saved state. Successful materials, ready attachments and original identifiers are preserved. Unknown sends are never repeated automatically.
Set the boundaries
Select “Recover failed stages within owner limits”: 0–5 attempts per job, 1–50 recoveries per day and 0–200 new generation calls. 0 attempts disables recovery; 0 calls permits only work using saved results.
Time changes need permission
The original publication time is kept by default. You may allow up to 1440 minutes of delay for a local publication waiting on media. Reconnecting accounts and unknown provider outcomes require a human.
Follow the recovery journal
The cabinet shows queue state, recovery outcome, cause and new generation call count. Optional automatic recovery of safe failures works without an open chat. Limits reset at 00:00 UTC; credit packs are never purchased. Handoff to a platform is not publication confirmation.
Access duration and limits, in one place
- Connection duration
- 1–90 days
- Materials and publications per day
- 1–200
- Daily caps reset
- 00:00 UTC
Separately for each connection
This time is fixed and does not follow your account time zone.
Defaults: 30 days, 20 materials, 20 publications and 20 replies per UTC day. Kontentika generation is disabled (0). Autopilot may remain connected until revoked.
Permissions and boundaries
You stay in control of access
- Choose the workspace, accounts, permissions and limits. Autopilot requires separate confirmation. Social sign-in happens on the platform’s own site.
- Only reading is selected by default. Autonomous publishing needs Autopilot, a separate permission and owner confirmation. Work and access can be stopped.
- You may still need to complete initial social sign-in, renewed provider consent and security checks. The agent receives no social passwords, cannot change payment details or administer users. Advertising requires separate permissions and a financial grant.
- Ask the agent to pause work. It can stop its local waiting jobs while keeping read access. Revoke the connection completely in AI agents. An operation already handed to a social platform may still finish; inspect its actual status.
- Connection keys are stored as hashes; social passwords and access tokens are never given to the agent.
What does it cost?
Ready text, image and video imports use zero Kontentika AI credits. Our generation uses applicable subscription allowances. Material/publication limits still apply, and your external agent may bill separately.
View plans ↗Before you begin
Common questions
Can I work only through the agent?
After initial setup, yes, for supported content, publishing, audience and analytics operations. Payments and provider access/security checks remain with the owner.
What costs money?
Ready text, image and video imports use zero Kontentika AI credits. Our generation uses applicable subscription allowances. Material/publication limits still apply, and your external agent may bill separately.
What does a daily limit mean?
It caps accepted actions during a UTC day: materials added, publications admitted, replies and generation. Subscription allowances also apply. The agent can read remaining allowance before starting.
What happens when I pause?
New mutations are paused and local waiting jobs are cancelled. Resume permits new work; it does not restart cancelled jobs or retract published posts.
Will my AI assistant work with Kontentika?
It needs remote MCP over Streamable HTTP with OAuth or a Bearer token, or the ability to call an HTTP API. Support depends on the assistant and its plan.
Let your agent help you connect
Copy this text and send it to your AI agent. It can find the guide and explain how to connect in your client.
View the instruction text
For developers
Two transports for the same set of actions. These settings are needed only for manual client configuration.
MCP over Streamable HTTP provides remote tools. OAuth with PKCE lets the owner sign in through Kontentika without handing over a key; dynamic client registration is supported.
A Bearer token is a secret connection key. Send it only in the Authorization header. The HTTP API exposes the same actions without MCP.
MCP: https://app.kontentika.app/mcp
HTTP: POST https://app.kontentika.app/api/agent/v1/{tool_name}
Authorization: Bearer <secret>Ready to connect your agent?
Connect your AI agent once, then manage social work by talking to it. It can maintain brand knowledge, prepare content, schedule posts, reply to your audience and read analytics. Autopilot removes per-post review; you choose the accounts, permissions and limits.
Connect an agent ↗