Agency SwarmvsCersei
Generated from the two spec rows. Green marks the better value where a spec has a clear direction. Everything else is just different.
Agency Swarm
VRSEN · Agent framework
OSS
Panel
6.31 spec wins
- Reliability
- 5.8
- Usefulness
- 6.2
- Cost
- 7.2
- Longevity
- 5.8
“You can appoint an agent CEO, which is the first time an org chart has been under version control.”
Cersei
Adib Mohsin · Agent framework
OSSMCP
Panel
6.27 spec wins
- Reliability
- 5.7
- Usefulness
- 6.0
- Cost
- 8.3
- Longevity
- 4.7
“It reads Claude Code's skills files and writes Claude Code's session files, so the migration path is helpfully signposted in one direction.”
Spec by spec
| Spec | Agency Swarm | Cersei |
|---|---|---|
| Architecture | ||
| Category | Agent framework | Agent framework |
| Runs | local | local |
| Platforms | macos, linux, windows | macos, linux |
| Context window | not documented | not documented |
| Protocols | ||
| MCP client | No | Yes |
| MCP server | No | No |
| Capabilities | ||
| Runs terminal commands | No | Yes |
| Multi-file edits | No | Yes |
| Git operations | No | YesAbstract ships /commit, /review and /diff slash commands over the working tree. |
| Browser control | No | YesWeb tools are among Abstract's 34 built-in tools; no browser automation loop is documented. |
| Sandboxed execution | No | No |
| Multi-agent orchestration | Yes | Yes |
| Headless / CI mode | No | YesAbstract offers --no-permissions with NDJSON output as a CI mode. |
| Models | ||
| Backbone | OpenAI, Anthropic, Google, xAI, Azure OpenAI, OpenRouter | Anthropic, OpenAI, Ollama, Azure OpenAI, vLLM |
| Bring your own model | YesOpenAI models are native; Anthropic, Gemini, Grok, Azure OpenAI and OpenRouter are reached through the LiteLLM router. | Yes |
| Local models | No | Yes |
| Cost | ||
| Pricing model | byok | byok |
| Starts at | $0/mo | $0/mo |
| Free tier | Yes | Yes |
| Bring your own key | Yes | Yes |
| Openness | ||
| Open source | Yes | Yes |
| License | MIT | MIT |
| GitHub stars | 4,587 | 460 |
Which one would each critic pick
| Critic | Agency Swarm | Cersei | Pick |
|---|---|---|---|
| El Juez | — | — | not enough reviews |
| El Amigo | 6.8 | 6.8 | no preference |
| El Crítico | 6.0 | 5.8 | Agency Swarm — It is a layer on the OpenAI Agents SDK, so its ceiling is that SDK's ceiling and an upstream breaking change is your breaking change. |
| El Profesor | 6.3 | 6.0 | Agency Swarm — State survives sessions through explicit load and save callbacks, which is the correct place to put persistence, and no evaluation of the topology is published. |
| La Inversora | 5.3 | 5.8 | Cersei — One named author, 453 stars, no company and nothing to charge for: the acquisition risk is zero here because there is no asset an acquirer could buy. |
| La Jefa | 5.8 | 5.3 | Agency Swarm — Free as a dependency for sixty engineers, and the meter is our model contract, but there is no headless story so it never runs unattended in our pipelines. |
| El Hacker | 7.5 | 7.5 | no preference |
Picks are derived from each critic's own scores. Humans vote on matchups on the duels page.