# Timon

Local workspace for agent teams: coordinates existing coding agents, tasks, reviews and routines through a shared background service.

Canonical profile: https://ki-agentenkompass.de/en/agents/timon/
Editor: [Michael Meierhoff](https://ki-agentenkompass.de/en/about/)
Sources reviewed: 2026-11-06
Basis: provider documentation. Examples are editorial suggestions, not hands-on product tests.

## Overview

Timon is classified as a standalone product with its standard access.

- Provider: PUBLIKO
- Product family: Timon
- Profile type: Agent workspace with orchestration
- Environment: Desktop app for Windows, macOS and Linux; web interface on your own host
- Connections: Claude Code, Codex, Copilot, Grok, opencode, Pi, Hermes, MCP and Git
- Setup: Install Timon, sign in to an agent CLI and configure models, working folders and approvals for each role.
- Pricing model: Free individual access; separate model costs, Pro pricing pending

## Key features

### Roles in an agent team

Orchestrator, coder, reviewer and further roles distribute work among separately configured agents.

Source: https://github.com/ai-timon/timon

### Code, review and merge

Tasks run through a Kanban board and separate Git worktrees; different models can write and review.

Source: https://timon-ai.com/

### Memory with scope

Rules apply globally or by project, role or agent and are injected at startup as appropriate.

Source: https://docs.timon-ai.com/plugins/memory

### Background routines

A service holds working state. Schedules, webhooks and workflows enable recurring assignments.

Source: https://timon-ai.com/

### Browser and multiple chats

An integrated browser and side-by-side conversations support the agents’ work.

Source: https://github.com/ai-timon/timon

### Models and fallbacks

Existing CLI access and additional compatible endpoints can be combined; fallbacks respond to quota limits.

Source: https://docs.timon-ai.com/usage/missions

## Example workflows

### Implement a scoped development goal

- Your input: Repository, acceptance criteria and permissions for the participating roles.
- Possible result: Decomposed tasks with changes and a separate review.
- What to check: Check tests, diffs and consent before pushing or publishing.

### Recurring project maintenance

- Your input: A clear inspection task, schedule and scoped access.
- Possible result: Repeatable runs with findings and updated working rules.
- What to check: Review new rules for accuracy and appropriate scope.

## Requirements and limits

### Provide host and backends

Requirement: The background service requires a running computer or server and available agent access.

Source: https://docs.timon-ai.com/installation/background

### Check continuation after restart

Limitation: A conversation turn interrupted by a restart is not automatically replayed. Check prior actions, then continue with a message.

Source: https://docs.timon-ai.com/usage/missions

### Plan scope is still changing

Not yet confirmed: The website and repository differ on plugin and plan details. Pro and Cloud prices are not yet specified on the pricing page.

Source: https://timon-ai.com/pricing

### Network logging is observation

Limitation: Optional connection logging records only traffic passing through its proxy and blocks no connections.

Source: https://docs.timon-ai.com/usage/network-activity

## A good fit if …

- you want to coordinate several existing coding agents
- you want to manage roles, memory and tasks on your own host
- you want to assign coding and review to different models

## Less suitable if …

- you want a fully managed service without operating your own host
- you expect fully automatic recovery after every failure

Editorial assessment based on provider documentation. Not a product test.

## Usable AI models

Open choice

Timon combines existing agent CLIs and supports additional Anthropic- or OpenAI-Responses-compatible endpoints, including personal keys, local servers and LiteLLM. Models and fallbacks are configurable; protocol support and quotas depend on the backend. The technical documentation qualifies the README’s blanket subscription framing.

- Model families named: Anthropic (Claude), OpenAI (GPT), xAI (Grok), DeepSeek, Moonshot (Kimi) · Other models can be connected freely
- Model choice: Automatic or your choice
- Your own API key: Yes
- Router or custom endpoint: Yes
- Local models: Yes
- Depends on: Plan, Admin approval

Documented · Sources accessed on: 2026-11-06

- [Timon: Add a provider](https://docs.timon-ai.com/usage/add-a-provider)

- [Timon: Models and fallbacks](https://docs.timon-ai.com/usage/missions)

## Europe & regional availability

Local operation and external model processing are assessed separately.

Reviewed scope: Your own Timon host and chosen model providers
Reviewed on: 2026-11-06

### Determine data region through the backend

With requirements

Timon can run on your computer or server. The host location does not determine processing by connected model providers; check their country access and data region separately.

Region covered: EU/EEA, United Kingdom and Switzerland

Documented reason: Custom compatible endpoints are supported; the selected provider determines external processing.

Official source: https://docs.timon-ai.com/usage/add-a-provider
Official source: https://docs.timon-ai.com/installation/background

## Accounts & permissions

Based on provider documentation. Permissions depend on the account, integration and organization policies. Review the permissions actually requested before use.

Configuration covered: Timon account, backend access and agent permissions
Sources reviewed: 2026-11-06

### Account used

Documented: Timon uses signed-in agent CLIs; additional provider access is configured with personal keys.

Evidence: https://github.com/ai-timon/timon
Evidence: https://docs.timon-ai.com/usage/add-a-provider

### Admin approval

Documented: Agents have approval modes and allowed or blocked tools. Orchestration permissions determine who may assign work to whom.

Evidence: https://docs.timon-ai.com/usage/agents

### Data access & team sharing

Documented: Keys can be scoped to a backend and specific projects or agents. The chosen CLI access imposes further boundaries.

Evidence: https://docs.timon-ai.com/usage/add-a-provider

## Prices and included usage

Pricing checked: 2026-11-06

The Free plan is intended for one person on one machine. Model access and hosting are supplied separately.

### Free

USD 0 for Timon

Agents, Kanban, memory, routines and browser according to the pricing page.

Price source: https://timon-ai.com/pricing

Pro and Cloud are announced; specific prices are pending. The listed Cloud plan runs on a server of your choice.

The repository and website differ on plugin scope. Check current license and plan terms before procurement.

Pricing details: https://timon-ai.com/pricing

## 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: Your own Timon service with external agent backends
Sources reviewed: 2026-11-06

### Files and working state

Documented: The service holds local state; backups and trash have their own retention rules.

Evidence: https://docs.timon-ai.com/usage/settings

### Sign-ins

Documented with conditions: CLI sign-ins and additional provider keys must remain valid.

Evidence: https://docs.timon-ai.com/usage/add-a-provider

### Starting further runs

Documented: Schedules and webhooks can start new tasks on the running host.

Evidence: https://timon-ai.com/

### Human intervention

Documented with conditions: Approval mode and tool rules determine when a question is required.

Evidence: https://docs.timon-ai.com/usage/agents

### Detecting errors

Documented: The mission view and conversation history show interrupted steps and errors.

Evidence: https://docs.timon-ai.com/usage/missions

### Resuming after failures

Documented with conditions: Conversation turns after service restarts require a new message; quota failures in Kanban tasks have separate retry rules.

Evidence: https://docs.timon-ai.com/usage/missions

### Editorial assessment

Editorial assessment: suitable for supervised agent teams with clear assignments; runtime, approvals and actual recovery should be tested in your own workflow.

## Memory and conversation continuity

Scope: Agent and project memory, rules and mission history
Sources accessed: 2026-11-06

Local memories with four scopes, visible rules and backend-dependent conversation continuation.

### Storage and scope

Documented: Timon keeps agent and project files plus rules and extracted facts in a local database. The service manages working state.

- [Timon: Architecture](https://github.com/ai-timon/timon)
- [Timon: Memory queries](https://github.com/ai-timon/timon/blob/6cdc82d4a0fe2f153c1ebd9d106b43a4d2827d29/orchestrator/src/memory-queries.ts)

### Selection and timing

Documented: Facts are captured after completed or failed missions. A nightly curation pass condenses them into rules; rules can also be added explicitly.

- [Timon: Memory capture](https://github.com/ai-timon/timon/blob/6cdc82d4a0fe2f153c1ebd9d106b43a4d2827d29/orchestrator/src/memory-capture.ts)
- [Timon: Memory synthesis](https://github.com/ai-timon/timon/blob/6cdc82d4a0fe2f153c1ebd9d106b43a4d2827d29/orchestrator/src/memory-synthesize.ts)
- [Timon: Memory API](https://docs.timon-ai.com/plugins/memory)

### Inspect, edit and delete

Documented: Files and rules can be edited or deleted. Deleted agents remain in trash for 30 days by default; project facts and backups are managed separately. Rule deletions leave audit entries.

- [Timon: Memory routes](https://github.com/ai-timon/timon/blob/6cdc82d4a0fe2f153c1ebd9d106b43a4d2827d29/orchestrator/src/plugins-bundled/memory/routes.ts)
- [Timon: Backups and trash](https://docs.timon-ai.com/usage/settings)
- [Timon: Memory rule deletion audit](https://github.com/ai-timon/timon/blob/6cdc82d4a0fe2f153c1ebd9d106b43a4d2827d29/orchestrator/src/memory-queries.ts)

### Agent and channel boundaries

Documented: Rules apply globally, to a project, to a role or to an agent. A duplicated agent starts with empty memory.

- [Timon: Memory API](https://docs.timon-ai.com/plugins/memory)
- [Timon: Agents](https://docs.timon-ai.com/usage/agents)

### Retrieval

Documented: Startup injects a digest, applicable rules and recent pitfalls. Memory search combines full-text search for rules and text search for facts.

- [Timon: Injections](https://docs.timon-ai.com/usage/agents)
- [Timon: Memory queries](https://github.com/ai-timon/timon/blob/6cdc82d4a0fe2f153c1ebd9d106b43a4d2827d29/orchestrator/src/memory-queries.ts)

### Session reset and compaction

Documented: Compaction uses the same conversation or a new one with a summary, depending on the backend. After a service restart, an interrupted turn remains for inspection; a new message continues it where supported by the backend.

- [Timon: Missions and compaction](https://docs.timon-ai.com/usage/missions)

## Context

The background service requires a running computer or server and available agent access.

Official product documentation reviewed. Workflows and assessments are editorial; hands-on product tests are outside the scope of this directory.

## Open points

Requirements and limits, not yet confirmed:
- Plan scope is still changing

## Sources

- https://timon-ai.com/

- https://github.com/ai-timon/timon

- https://docs.timon-ai.com/usage/agents

- https://docs.timon-ai.com/usage/missions

- https://docs.timon-ai.com/plugins/memory

- https://docs.timon-ai.com/usage/settings

- https://docs.timon-ai.com/usage/add-a-provider

- https://timon-ai.com/pricing

- https://timon-ai.com/legal-notice

- https://docs.timon-ai.com/installation/background

- https://docs.timon-ai.com/usage/network-activity

- https://github.com/ai-timon/timon/blob/6cdc82d4a0fe2f153c1ebd9d106b43a4d2827d29/orchestrator/src/memory-capture.ts

- https://github.com/ai-timon/timon/blob/6cdc82d4a0fe2f153c1ebd9d106b43a4d2827d29/orchestrator/src/memory-synthesize.ts

- https://github.com/ai-timon/timon/blob/6cdc82d4a0fe2f153c1ebd9d106b43a4d2827d29/orchestrator/src/memory-queries.ts

- https://github.com/ai-timon/timon/blob/6cdc82d4a0fe2f153c1ebd9d106b43a4d2827d29/orchestrator/src/plugins-bundled/memory/routes.ts

## Provider

Timon is developed by PUBLIKO in France. The code is published under AGPL-3.0; the SDK and plugin templates have separate license terms.

- Operator: PUBLIKO
- Organization: Software provider
- Website: https://timon-ai.com/
- Reviewed: 2026-11-06

- Repository: https://github.com/ai-timon/timon
- GitHub-Stars: 50 (2026-11-06)
- License: AGPL-3.0

- Provider source: https://timon-ai.com/legal-notice

- Provider source: https://github.com/ai-timon/timon

[Sources and method](https://ki-agentenkompass.de/en/sources/)
